Upload
ngohuong
View
215
Download
0
Embed Size (px)
Citation preview
Instructions for use
Title Phase Resistance Feedback Control and Modeling of Thick SMA Actuators
Author(s) 李, 軍鋒
Citation 北海道大学. 博士(工学) 甲第11169号
Issue Date 2013-12-25
DOI 10.14943/doctoral.k11169
Doc URL http://hdl.handle.net/2115/54652
Type theses (doctoral)
File Information Junfeng_Li.pdf
Hokkaido University Collection of Scholarly and Academic Papers : HUSCAP
PHASE RESISTANCE FEEDBACK
CONTROL AND MODELING OF
THICK SMA ACTUATORS
DOCTORAL DISSERTATION
JUNFENG LI
DIVISION OF HUMAN MECHANICAL SYSTEM AND DESIGN
GRADUATE SCHOOL OF ENGINEERING
HOKKAIDO UNIVERSITY
DECEMBER, 2013
2
CLAIMS OF ORIGINALITY
i
CLAIMS OF ORIGINALITY
This doctoral thesis contains no material which has been accepted for the award of
any other degree or diploma in any university. To the best of the author's knowledge and
belief, it contains no material previously published or written by another person, except
where due reference is made in the text.
Signature: Junfeng Li
December 2013
ii
CONTENTS
iii
CONTENTS
CLAIMS OF ORIGINALITY ........................................................................................ i
CONTENTS ................................................................................................................ iii
DISSERTATION ABSTRACT .................................................................................... vii
LIST OF FIGURES ..................................................................................................... xi
LIST OF TABLES .................................................................................................... xvii
CHAPTER 1: INTRODUCTION .................................................................................. 1
1.1 Background ..................................................................................................... 1
1.2 Research Objectives and Approach .................................................................. 9
1.3 Outline of Thesis ........................................................................................... 10
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS ............................ 11
2.1 Phase Transformation Temperature of SMA .................................................. 13
2.2 Two Different Shape-Memory Effects ........................................................... 14
2.2.1 One-way of SMA.................................................................................... 16
2.2.2 Two-way of SMA ................................................................................... 16
2.2.3 Pseudo-elasticity ..................................................................................... 17
2.3 Micro and Macro Analysis of SMA ............................................................... 18
2.4 SMA Actuators .............................................................................................. 20
2.5 Literature Overview....................................................................................... 23
2.5.1 Modeling of SMA ................................................................................... 23
2.5.2 Controlling of SMA ................................................................................ 30
2.6 Chapter Summary .......................................................................................... 34
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR..... 37
CONTENTS
iv
3.1 Introduction ................................................................................................... 37
3.2 Motivation and Target .................................................................................... 37
3.3 Method .......................................................................................................... 38
3.4 Experimental Setup ....................................................................................... 40
3.5 Results........................................................................................................... 44
3.5.1 Results with Different Ambient Temperature .......................................... 44
3.5.2 Results with Binary Control .................................................................... 47
3.6 Consideration and Discussion ........................................................................ 52
3.7 Chapter Summary .......................................................................................... 56
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED ................................................................................................... 57
4.1 Introduction ................................................................................................... 57
4.2 Motivation and Target .................................................................................... 59
4.3 Method .......................................................................................................... 60
4.3.1 Phase Resistance ..................................................................................... 60
4.3.2 Phase Resistance Feedback Control Method (PRFC) .............................. 62
4.3.3 PID controller ......................................................................................... 64
4.3.4 Experimental Setup ................................................................................. 66
4.4 Results........................................................................................................... 70
4.4.1 Phase Resistance Identification ............................................................... 70
4.4.2 Tuning the PID Parameters ..................................................................... 74
4.4.3 Results with PRFC .................................................................................. 77
4.5 Consideration and Discussion ........................................................................ 81
4.6 Chapter Summary .......................................................................................... 86
CONTENTS
v
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK .......................................... 89
5.1 Introduction ................................................................................................... 89
5.2 Motivation and Target .................................................................................... 89
5.3 Method .......................................................................................................... 90
5.3.1 Phase Resistance with Displacement Feedback Control (PRDFC) ........... 90
5.3.2 Control System ....................................................................................... 93
5.4 Results........................................................................................................... 95
5.4.1 Tuning PID Parameters ........................................................................... 95
5.4.2 Results with PRDFC ............................................................................... 98
5.5 Consideration and Discussion ...................................................................... 102
5.6 Chapter Summary ........................................................................................ 105
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL.................................................................................................... 107
6.1 Introduction ................................................................................................. 107
6.2 Motivation and Target .................................................................................. 108
6.3 Method ........................................................................................................ 108
6.3.1 Thermal Model of Heat Transfer and Temperature ................................ 108
6.3.2 Phase Transformation and Mechanical Model ....................................... 111
6.4 Results......................................................................................................... 118
6.5 Consideration and Discussion ...................................................................... 125
6.6 Chapter Summary ........................................................................................ 127
CHAPTER 7: CONCLUSIONS AND FUTURE WORKS ........................................ 129
7.1 Conclusions ................................................................................................. 129
CONTENTS
vi
7.2 Future Works ............................................................................................... 131
ACKNOWLEDGEMENTS ...................................................................................... 133
APPENDIX A ........................................................................................................... 135
a. Binary Control Code with Matlab ................................................................ 135
b. PID Controller Code with Resistance as Feedback ....................................... 137
c. Microcontroller Code .................................................................................. 138
APPENDIX B ........................................................................................................... 143
a. Microcontroller ........................................................................................... 143
b. Power Source, Displacement and Force Sensor ............................................ 144
REFERENCES ......................................................................................................... 149
DISSERTATION ABSTRACT
vii
DISSERTATION ABSTRACT
Title of dissertation submitted for the degree
Phase resistance feedback control and modeling of thick SMA Actuators
Shape memory alloy (SMA) actuators have great potential in niche applications
where space, weight, cost and noise are crucial factors. These applications include
mobile robots, microrobot manipulation, smart structures, and artificial muscles.
Despite many of the advantages, they remain mostly as experimental actuators due to
their perceived slow response speed, low accuracy and controllability. In the past,
research had focused on position and force control of thin SMA wires, because they can
be cooled fast in air and as it is easy to obtain rapid response speed for SMA actuators
in this manner. Due to hysteresis and significant nonlinearities in the behavior of shape
memory alloy actuators, it is difficult to obtain rapid response speed of SMA actuators,
which have limited the application of these actuators, especially for thick SMA wires. In
this thesis, effective control systems are applied to achieve rapid response speed control
of SMA wire actuators.
In this thesis, experimental tests are conducted in chapter 3 to show the existence of
latency duration during the heating and cooling process for thick SMA wires 0.5mm in
diameter. The heating times (5s for SMA1 and 3s for SMA2) are decided to obtain the
latency duration with ‘on-off’ binary control. It is important to avoid overheating of
SMA wires which leads to long latency duration and slow to cool when the power is
DISSERTATION ABSTRACT
viii
turned off. In addition, the experimental results show that the ambient temperature has
an effect on the cooling speed. Therefore, the ambient temperature needs to be stable
when the experiments conducted.
Then, an approach is proposed in chapter 4 to design and control a thick SMA
actuator to achieve rapid response speed control with two connected SMA wires. In the
proposed method, a structure with two connected SMA wires is designed and then the
concept of phase resistance is defined to use it as feedback in the response speed control
system. Phase resistance feedback control (PRFC) minimizes cooling time by
shortening the long latency duration of thick SMA wires. To accurately identify phase
resistances, experiments showed that it is important to determine the major hysteresis
loop. Experimental results that demonstrate the advantages and justify the concepts are
also presented.
Subsequently, another method is proposed in chapter 5, the phase resistance with
displacement feedback control (PRDFC), combing both the phase resistance and
displacement as feedback, minimizes cooling time by shortening the long latency
duration of thick SMA wires. PRDFC using segmented SMA wires shortens the latency
duration of SMA wire, which coordinates with each other to make sure the continuity of
output displacement. Two sets of experiment are tested using the step and ramp signals
as reference input. Experimental results show that rapid response speed is achieved
using this method in comparison with the case in which only displacement is used as
feedback.
In chapter 6, a successful empirical relation is proposed in order to model the major
and minor hysteresis loops of behavior for SMA actuators, which considers the amount
of the austenite fraction transformed at a temperature based on Liang and Rogers model.
DISSERTATION ABSTRACT
ix
Finally, the conclusions of the whole work are described and the future works are
presented in chapter 7.
x
LIST OF FIGURES
xi
LIST OF FIGURES
1-1 The structure and the schematic of bending movement of biomimetic fin [1] .... 2
1-2 Micro-robot fish prototype [1] .......................................................................... 3
1-3 Endoscope actuated by SMA wires [2] ............................................................. 4
1-4 An illustration of the mechanical setup with high voltage [4] ............................ 5
1-5 Strain-temperature characteristics [5] ............................................................... 8
1-6 (i) Peltier sandwich structure; (ii) Photo of experiment setup of SMA actuator
[5] ................................................................................................................... 8
2-1 Schematic of austenite fraction-temperature hysteresis ................................... 14
2-2 Schematic for one way of SMA ...................................................................... 15
2-3 Schematic for two ways of SMA .................................................................... 15
2-4 Schematic of stress VS temperature ................................................................ 17
2-5 Schematic of micro and macro phenomenon of SMA ..................................... 19
2-6 Schematic of different SMA actuators ............................................................ 22
2-7 Play hysteresis operator .................................................................................. 28
2-8 Generalized play hysteresis operator .............................................................. 29
3-1 Schematic of output displacement for SMA wire ............................................ 38
3-2 Schematic of the binary control with latency duration Tcd .............................. 39
3-3 Schematic of experimental setup .................................................................... 40
3-4 SMA wires used in the experiment ................................................................. 41
3-5 SMA wires used in the experiment ................................................................. 42
3-6 (i) Photo of the SMA actuator and (ii) Photo of experimental setup ................ 43
3-7 10s as heating time for SMA .......................................................................... 45
LIST OF FIGURES
xii
3-8 Output displacement for SMA ........................................................................ 46
3-9 Results with different ambient temperature for SMA ...................................... 46
3- 10 Heating time of SMA1 ................................................................................ 48
3- 11 Output displacement of SMA1 in ambient temperature 24°C ....................... 48
3-12 Binary control with latency duration Tcd for SMA1 in ambient temperature
24°C .............................................................................................................. 49
3- 13 Output displacement of SMA2 in ambient temperature 24°C ....................... 50
3-14 Binary control with latency duration Tcd for SMA2 in ambient temperature
24°C .............................................................................................................. 50
3-15 Binary control with different heating voltage for SMA1 in ambient
temperature 24°C........................................................................................... 54
3-16 (i) SMA wires; (ii) Results for SMA 0.15mm in diameter; (iii) Results for
SMA 0.5mm in diameter in ambient temperature 21°C .................................. 55
4-1 An illustration of variable structure control [74] ............................................. 58
4-2 Schematic of the binary control with latency duration Tcd ............................... 59
4-3 Phase transformation resistances VS strain ..................................................... 61
4-4 Schematic for the PRFC method .................................................................... 63
4-5 Output displacement with the PRFC ............................................................... 64
4-6 Block diagram of PRFC ................................................................................. 66
4-7(i) Schematic outline of the experimental setup and (ii) Photo of the
experimental setup ......................................................................................... 67
4-8 Driver circuit .................................................................................................. 68
4-9 Control loop of the experiment ....................................................................... 68
4-10 Results with different cut-off frequency for SMA1 and SMA2 in ambient
LIST OF FIGURES
xiii
temperature 24°C........................................................................................... 69
4-11 Results of resistance with heating time 5s for SMA1 in ambient temperature
24°C .............................................................................................................. 71
4-12 Results of resistance with heating time 3s for SMA2 in ambient temperature
24°C .............................................................................................................. 71
4-13 Major hysteresis loop for SMA1 in ambient temperature 24°C ..................... 72
4-14 Major hysteresis loop for SMA2 in ambient temperature 24°C ..................... 72
4-15 Results with different PID parameters for SMA1 in ambient temperature 24°C
...................................................................................................................... 75
4-16 Results with different PID parameters for SMA2 in ambient temperature 24°C
...................................................................................................................... 76
4-17 Results with PRFC for SMA1 in ambient temperature 24°C ......................... 78
4-18 Results with PRFC for SMA2 in ambient temperature 24°C ......................... 79
4-19 Total output displacement with PRFC in ambient temperature 24°C ............. 80
4-20 Detailed output displacement with two cycles for PRFC in ambient
temperature 24°C........................................................................................... 80
4-21 Results with the PRFC and binary control for SMA1 in ambient temperature
24°C .............................................................................................................. 82
4-22 Results with the PRFC and binary control for SMA2 in ambient temperature
24°C .............................................................................................................. 82
4-23 Binary control with or without resistance calculation in ambient temperature
24°C .............................................................................................................. 84
4-24 Results with the binary control for 5 tests in ambient temperature 24°C ....... 85
5-1 Schematic of step reference for the PRDFC method ....................................... 91
LIST OF FIGURES
xiv
5-2 Schematic of ramp reference for the PRDFC method ..................................... 92
5-3 Block diagram of PRDFC, (i) Displacement feedback control; (ii) Phase
resistance feedback control ............................................................................ 93
5-4 Schematic of the experimental setup for displacement feedback control ......... 94
5-5 Results with different parameters of PI controller in ambient temperature 24°C
...................................................................................................................... 96
5-6 Input voltage for different parameters of PI controller in ambient temperature
24°C .............................................................................................................. 96
5-7 Results of resistance for SMA1 with PRDFC in ambient temperature 24°C;
resistance control: Kp=4000, Ki=0.2 ............................................................... 97
5-8 Results of output displacement for SMA1 with PRDFC in ambient temperature
24°C; resistance control: Kp=4000, Ki=0.2 ..................................................... 98
5-9 Results with PRDFC and traditional method in ambient temperature 24°C;
displacement control: Kp=1000, Ki=0.2; resistance control: Kp=4000, Ki=0.2 . 99
5-10 Results of resistance for SMA2 with PRDFC; displacement control: Kp=1000
and Ki=0.2 ..................................................................................................... 99
5-11 Results with PRDFC and traditional method in ambient temperature 24°C;
displcement control: Kp=1000 and Ki=0.2; resistance control: Kp=4000, Ki=0.2
.................................................................................................................... 100
5-12 Results of resistance for SMA1 with PRDFC in ambient temperature 24°C;
resistance control: Kp=4000, Ki=0.2 ............................................................. 100
5-13 Results of resistance for SMA2 with PRDFC in ambient temperature 24°C;
resistance control: Kp=1000, Ki=0.2 ............................................................. 101
5-14 Results with overheating for SMA1 in ambient temperature 24°C;
LIST OF FIGURES
xv
displacement control: Kp=1000 and Ki=0.2; resistance control: Kp=4000 and
Ki=0.2 .......................................................................................................... 104
6-1 Block diagram model of SMA ...................................................................... 109
6-2 Schematic of the input voltage ..................................................................... 110
6-3 Schematic of the output temperature ............................................................ 111
6-4 Schematic of martensite fraction-temperature hysteresis ............................... 112
6-5 Schematic of typical austenite fraction-temperature hysteresis ..................... 115
6-6 Schematic of austenite fraction-temperature hysteresis with modification ..... 115
6-7 Schematic of the input voltage ..................................................................... 119
6-8 Schematic of the displacement VS input voltage .......................................... 121
6-9 Simulated austenite fraction versus time ....................................................... 121
6-10 Simulated austenite fraction versus temperature ......................................... 122
6-11 Curve for fitting the normalized martensite starting temperature ................. 122
6-12 Curve for fitting the martensite starting temperature ................................... 123
6-13 Simulation results of the output displacement VS the input voltage ............ 123
6-14 Plot with experimental and simulated data .................................................. 125
7-1 Block diagram for the compensation based on the inverse model ................. 131
B-1 Microcontroller used in experiment ............................................................. 143
B-2 Power source used in experiment ................................................................. 145
B-3 Displacement sensor used in experiment ...................................................... 146
B-4 Force sensor used in experiment .................................................................. 147
xvi
LIST OF TABLES
xvii
LIST OF TABLES
3-1 Parameters for the experiments....................................................................... 42
3-2 Latency duration with different ambient temperature ...................................... 45
3-3 Latency duration with different heating time for SMA1 .................................. 47
3-4 Latency duration with different heating time for SMA2 .................................. 51
3-5 Heating time and cycle time of SMA wires..................................................... 51
4-1 List of important data in Fig. 4-3 .................................................................... 61
4-2 Transformation resistances parameters ........................................................... 70
4-3 Parameters of the SMA1 and SMA2 ............................................................... 81
6-1 Parameters of the experiments ...................................................................... 119
6-2 Simulation parameters .................................................................................. 124
B-1 Detailed information for microcontroller ..................................................... 144
xviii
CHAPTER 1: INTRODUCTION
1
CHAPTER 1:
INTRODUCTION
1.1 Background
A robot is a mechanical or virtual artificial agent, usually an electro-mechanical
machine that is guided by a computer program or electronic circuitry. Robots can be
autonomous or semi-autonomous. By mimicking a life like appearance or automating
movements, a robot may convey a sense of intelligence or thought of its own. Robots
have replaced humans in the assistance of performing those repetitive and dangerous
tasks which humans prefer not to do, or are unable to do due to size limitations, or even
those such as in outer space or at the bottom of the sea where humans could not survive
the extreme environments. There are many kinds of robots, including mining robots,
military robots, teleoperated robots and so on.
Robotics is the branch of technology that deals with the design, construction,
operation, and application of robots, as well as computer systems for their control,
sensory feedback, and information processing. These technologies deal with automated
machines that can take the place of humans in dangerous environments or
manufacturing processes, or resemble humans in appearance, behavior, and/or cognition.
Many of today’s robots are inspired by nature contributing to the field of bio-inspired
robotics. Actuators are like the “muscles” of a robot, the parts which convert stored
energy into movement. By far the most popular actuators are electric motors that spin a
wheel or gear, and linear actuators that control industrial robots in factories. But there
CHAPTER 1: INTRODUCTION
2
are some recent advances in alternative types of actuators, powered by electricity,
chemicals, or compressed air.
An actuator is a type of driving mechanism for moving or controlling a mechanism or
system. It is operated by a source of energy, usually in the form of an electric current,
hydraulic fluid pressure or pneumatic pressure, and converts that energy into some kind
of motion. An actuator is the mechanism by which a control system acts upon an
environment. The control system can be simple (a fixed mechanical or electronic
system), software-based (e.g. a printer driver, robot control system), or a human or other
agent. As control and robotic systems continue to decrease in size and weight, there has
been a continuing trend in technology towards ever-smaller scales for mechanical,
optical as well as electro-mechanical devices. The actuator must therefore undergo
similar miniaturisation in design and construction.
Fig. 1-1 The structure and the schematic of bending movement of biomimetic fin [1]
CHAPTER 1: INTRODUCTION
3
Fig. 1-2 Micro-robot fish prototype [1]
Following this trend, factors such as power consumption, work density, costs and
space constraints gain increased importance in the selection of suitable technologies.
However, conventional actuators, including electric motors, pneumatic and hydraulic
actuators, suffer a large reduction in power that they can deliver as they are scaled down
in size and weight. This constraint has led to the emergence and development of novel
actuator technologies such as piezoelectric actuators, electrostatics, magnetostrictive
materials and shape memory alloys (SMA). Wang et al. proposed a fish robot actuated
by Shape Memory alloys [1], Fig. 1-1 and Fig. 1-2 show the details of fish robot.
Shape memory alloys (SMA) are metallic alloys which deform at low temperatures
and return to the original undeformed state when heated to higher temperatures. The
shape memory effect is a consequence of a reversion in the crystalline structure between
the low temperature and high temperature phases, which are respectively called the
CHAPTER 1: INTRODUCTION
4
martensite and the austenite of the SMA. The martensite phase is nonsymmetric and
relatively soft, while the austenite phase is symmetric and relatively hard and has a
much higher Young's Modulus. Heating the SMA can be done via Joule heating, which
is resistively heating the material using electric current. Of all the SMAs that have been
discovered so far, NiTi shape memory alloys, also known as Nitinol, have proven to be
the most flexible and successful in engineering applications. One of the ways SMAs are
commonly used is in the form of wires. Already, SMA have been used in a variety of
actuation applications because of advantages such as excellent power-to-mass ratios,
reliability, and silent actuation, such as endoscope (Fig. 1-3).
Fig. 1-3 Endoscope actuated by SMA wires [2]
Ikuta, Tsukumoto and Hirose proposed a control system for a shape memory alloy
(SMA) servo actuator, and its application to a unique medical tool [2]. It is thought that
the electric resistance value of an SMA can be utilized to monitor the transformation of
the SMA directly. Therefore, an antagonistic transformation control scheme using
CHAPTER 1: INTRODUCTION
5
electric resistance feedback is proposed and is verified by several experiments. The
bending angle of each segment can be controlled individually from outside by using
antagonistic electric resistance feedback without any motion sensors. Featherstone and
Teh proposed a method for improving the speed of actuators based on SMA by
increasing the rate at which an SMA element can safely be heated [3]. The method
consists of measuring the electrical resistance of an SMA element, calculating a
maximum safe heating current as a function of measured resistance, and ensuring that
the actual heating current does not exceed this maximum value. In effect, resistance is
being used as a form of temperature measurement, and the maximum safe heating
current is designed to prevent overheating. Experimental results show a substantial
increase in the maximum velocity attainable by this robot, without any change in the
cooling regime, purely as a result of faster heating.
Fig. 1-4 An illustration of the mechanical setup with high voltage [4]
When the SMA wire is applied in fast moving robots, the response speed of the
actuator is important, especially for robots which need thick SMA wires to provide large
forces. However, thick SMA wires are accompanied by large latency durations which
CHAPTER 1: INTRODUCTION
6
slow down the reaction speed of the SMA actuators. Due to the slow speed caused by
hysteresis phenomena in SMA which limits the practical application as actuators, many
researches had focused on improving the response speed of the SMA.
Vollach and Shilo explored the capabilities of a fast one-directional actuation mode
based on one-occasional rapid Joule heating of SMA elements [4]. For this purpose, a
unique experimental system (Fig. 1-4) has been developed that applies a high-voltage
electric pulse to a detwined NiTi wire and measures the resulting displacement due to
the martensite to austenite phase transformation. The research demonstrates the great
potential of SMA for applications that require high speeds and large displacements
one-occasional actuation. However, this method demands maximum current (300 A) to
heat the SMA, which is not practically available for application.
As shown in Fig. 1-5, Selden, Cho and Asada proposed the definition of state
transition threshold temperatures and calculated the latency duration when the
temperature increases from TC to TCH and decreases from TH to THC [5]. In steady of
investigation the relationship between latency duration and phase transformation
temperature of SMA, the relationship the latency duration and output displacement is
investigated in chapter 3.
In addition, a method eliminated the latency duration associated with phase transition
of SMA actuators in proposed. As shown in Fig. 1-6, this control method is
implemented using the Peltier effect thermoelectric devices for selective
segment-by-segment heating and cooling. However, due to thermal conductivity along
the wire, it might be troublesome to regulate adjacent segments at different temperatures.
Then, segments are heated and cooled to extreme temperature. there is heat transfer
between adjacent segments. In addition, one potential limitation to the experimental
CHAPTER 1: INTRODUCTION
7
apparatus built with Peltier effect is that segments of a SMA wire may shift to adjacent
units, as the SMA wire shrinks and expands. This may cause some error when an
adjacent segment at a different thermal state is brought to the neighboring segment. The
worst case scenario is that every pair of adjacent segments takes different thermal states.
Another potential limitation is that more energy will be needed to heat SMA since the
Peltier needs to be heated first compared with heating the SMA by Joule heating. At last,
the experimental apparatus is heavy because of many Peltier models in the apparatus.
However, these limitations caused by the Peltier models can be eliminated using
electricity to heat the SMA actuators, which are demonstrated in chapter 4. Theoretically,
both latency durations can be shortened which is demonstrated. However, due to the
electricity is used to heat the SMA wire in this thesis, the latency duration when
temperature increases from martensite finish to austenite start is short compared with
the latency duration when the SMA is cooled by air and temperature decreases from
austenite finish to martensite start. Only the latency duration when temperature
decreases from austenite finish to martensite start is shortened by the proposed method.
Other researches about achieving fast response speed and using resistance as
feedback will be introduced in chapter 4 as well. The detailed information about how to
implement Joule heat to shorten the latency duration during the position control using
phase resistance and displacement as feedback is discussed in chapter 5.
CHAPTER 1: INTRODUCTION
8
Fig. 1-5 Strain-temperature characteristics [5]
Fig. 1-6 (i) Peltier sandwich structure;
(ii) Photo of experiment setup of SMA actuator [5]
(i)
(ii)
CHAPTER 1: INTRODUCTION
9
1.2 Research Objectives and Approach
Due to hysteresis effect and nonlinear relationship, SMA actuators have generally
been considered to be slow, inaccurate and difficult to control continuously. The
primary objective of this research is to demonstrate that more rapid response speed of
SMA actuators can be achieved compared with the conventional method using new
design and implementation of practical and effective control systems.
Two approaches proposed here that have been adopted for achieving rapid response
speed. One is phase resistance feedback control (PRFC), which uses two SMA wires
and the results with PRFC will be compared with that of traditional control method
using only one SMA wire applied in PRFC. The other is phase resistance with
displacement feedback control (PRDFC), which uses two SMA wires as well and the
results with PRDFC will be compared with that of traditional control method using one
SMA wire. The total length of SMA wire used in both experimental tests for the second
approach is the same.
In order to apply the phase resistance as feedback to obtain rapid response speed, a
SMA actuator using two SMA wires which are connected together using insulation joint
to prevent short circuiting is proposed. Then, phase transformation resistances of SMA
wires are identified using experimental data. The critical aspect of the proposed method
is to shorten the latency time caused by hysteresis effect of SMA actuators [5]. The
SMA wires are heated separately to prevent interruption arisen from power. The heating
and cooling processes depends on the phase resistances of both wires. And experimental
tests are conducted to testify the effectiveness of the proposed methods.
CHAPTER 1: INTRODUCTION
10
1.3 Outline of Thesis
This thesis is divided into seven chapters and organized in the following manner.
In chapter 2, some background information on SMAs, including more detailed
descriptions of their phases and the phase transformations, as well as the various
arrangements illustrating how SMAs are used as actuators.
In chapter 3, an experimental setup is built in order to test the latency durations of
SMA wires. Detailed information about the experimental setup is demonstrated and the
results of two SMA wires show whether the latency duration exists during both the
heating and cooling processes and the ambient temperature has an effect on the cooling
speed or not. In addition, a criteria is decided to determine the martensite start
displacement during the cooling process.
In chapter 4 and chapter 5, PRFC and PRDFC are proposed here to achieve rapid
response speed of SMA actuator with two connected SMA wires, respectively. The basic
theory of PRFC and PRDFC are introduced and the control systems of PRFC
demonstrate how they work using phase resistances as feedback. Results with PRFC
and PRDFC are compared with that of binary control and traditional control method.
In addition, in chapter 6, a model is proposed to try to represent the major and minor
hysteresis loops of behavior for SMA actuators. The simulation as well as the
experimental results will be presented and compared.
Finally, in chapter 7, a summary of the research achievements will be provided
together with some discussion of future works
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
11
CHAPTER 2:
NICKEL-TITANIUM SHAPE MEMORY
ALLOYS
In the 1930s, the first reported steps towards the discovery of the shape-memory
effect were taken. According to Otsuka and Wayman, A. Ölander discovered the
pseudoelastic behavior of the Au-Cd alloy in 1932. Greninger and Mooradian (1938)
observed the formation and disappearance of a martensitic phase by decreasing and
increasing the temperature of a Cu-Zn alloy. A decade later, the basic phenomenon of
the memory effect governed by the thermoelastic behavior of the martensite phase was
widely reported by Kurdjumov and Khandros (1949) and also by Chang and Read
(1951).
In 1962-1963, the nickel-titanium alloys were developed by the United States Naval
Ordnance Laboratory and commercialized under the trade name Nitinol (an acronym for
Nickel Titanium Naval Ordnance Laboratories). Their remarkable properties were
discovered by accident. A sample that was bent out of shape many times was presented
at a laboratory management meeting. Muzzey, one of the associate technical directors,
decided to see what would happen if the sample was subjected to heat and held his pipe
lighter underneath it [6]. To everyone's amazement the sample stretched back to its
original shape.
There is another type of SMA, called Magnetic shape-memory alloys (MSMAs). Due
to martensitic phase transformation, MSMAs are ferromagnetic materials which exhibit
large strains under the influence of an applied magnetic field.
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
12
MSMAs, with near-stoichiometric Ni2MnGa being the most studied example, differ
from other magnetostrictive materials, such as Terfenol-D and Galfenol, as they produce
much larger strains by twinning, sometimes as large as 9%, under relatively low bias
magnetic fields. The mechanism is based on the magnetic anisotropy of the material. As
other SMA which change phase between austenite and martensite with the application
of thermal energy, MSMAs produce a similar phase transformation between martensite
1 and martensite 2 (the two variants). Few models have been developed which describe
the constitutive response of MSMAs. To describe the materials behavior,
thermodynamic modeling is typically used. A shift in the direction of magnetization is
produced when applying a stress to a fully strained element exposed to a bias field
because of the nature of MSMAs. The magnitude of this shift changes according to the
strength of the applied field and material properties. Using Faraday's law of induction,
MSMAs may be used for energy harvesting using a pickup coil, or inductor [7-8].
Shape-memory alloys are typically made by casting, using vacuum arc melting or
induction melting. These are special techniques used to keep impurities in the alloy to a
minimum and ensure the metals are well mixed. The ingot is then hot rolled into longer
sections and then drawn to turn it into wire. The way in which the alloys are “trained”
depends on the properties wanted. The “training” dictates the shape that the alloy will
remember when it is heated. This occurs by heating the alloy so that the dislocations
re-order into stable positions, but not so hot that the material recrystallizes. They are
heated to between 400°C and 500°C for 30 minutes. They are then shaped while hot and
are cooled rapidly by quenching in water or by cooling with air. The manufacturing
equipment can be found in many references, such as [9-10]. Concerning about
manufacturing routes of SMA, in the past, many research had focused on this field
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
13
[11-12].
2.1 Phase Transformation Temperature of SMA
The austenite phase transformations of the alloy can be characterized by four
transformation temperatures:
AST : Austenite start temperature
AFT : Austenite finish temperature
MST : Martensite start temperature
MFT : Martensite finish temperature
The martensite fraction can be used to represent the amount of martensite phase in the
alloy during the heating and cooling processes [13]. As shown in Fig. 2-1, with a
temperature less than MFT , the NiTi alloy consists only of the martensite phase. As the
temperature increases beyond AST , austenite begins to form in the alloy and when the
temperature exceeds AFT , the alloy is primarily in the austenite phase. As the alloy is
cooled, martensite begins to form when the temperature drops below MST , and when the
temperature reaches MFT , the alloy is again fully martensitic.
The transition between the austenite and martensite phases can be characterized by a
wide thermal hysteresis loop, especially for thick SMA wires. The hysteresis varies
according to the alloy system. For example, the temperature hysteresis is generally
between 30°C-50°C for NiTi alloys. During phase transitions between martensite and
austenite, most of the physical properties of SMAs vary, including Young's Modulus,
electrical resistance, heat capacity and thermal conductivity. In the possible range where
both martensite and austenite co-exist, nonlinearities and hysteresis are prominent, and
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
14
they are influenced by material composition, processing and the number of activated
cycles [14].
Fig. 2-1 Schematic of austenite fraction-temperature hysteresis
2.2 Two Different Shape-Memory Effects
In addition to common shape change effects such as elastic and plastic deformations,
as well as thermal expansion and contraction, SMA also exhibit three shape memory
characteristics. Two common effects are one-way and two-way shape memory. A
schematic of the effects is shown below.
As shown in Fig. 2-2 and Fig. 2-3, the procedures are very similar: starting from
martensite (a), adding a reversible deformation for the one-way effect or severe
deformation with an irreversible amount for the two-way (b), heating the sample (c) and
cooling it again (d).
Temperature
Au
sten
ite
frac
tion
0
100%
TMF TAS TMS TAF
Heating
Cooling
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
15
Fig. 2-2 Schematic for one way of SMA
Fig. 2-3 Schematic for two ways of SMA
One-way
Two-way
a
b
c
d
a
b
c
d
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
16
2.2.1 One-way of SMA
NiTi alloys exhibits the Shape Memory Effect (SME) when it is deformed while in
the martensitic phase and then unloaded while still at a temperature below MFT . If it is
subsequently heated above AFT , it will regain its original shape by transforming back
into the parent austenitic phase. When the metal cools again it will remain in the hot
shape, until deformed again. Since martensite variants have been reoriented by stress,
the reversion to austenite produces a large transformation strain having the same
amplitude but the opposite direction with the inelastic strain. With the one-way effect,
cooling from high temperatures does not cause a macroscopic shape change. A
deformation is necessary to create the low-temperature shape. On heating,
transformation starts at AST and is completed at AFT (typically 2°C to 20°C or hotter,
depending on the alloy or the loading conditions). AST is determined by the alloy type
and composition and can vary between -150°C and 200°C.
The above described phenomenon is called one-way shape memory effect (or simply,
shape memory effect) because the shape recovery is achieved only during heating.
2.2.2 Two-way of SMA
The two-way shape memory effect is less pronounced than the one-way effect and
usually requires training. In the shape memory effect discussed above, what is
remembered was the shape of the parent phase only; but, under certain conditions it is
possible to two different shapes: one at low temperatures, and one at the
high-temperature shape. This effect was first called ‘‘the reversible shape memory
effect’’ [15-16], but now it is called two-way shape memory effect.
SMA can be trained to exhibit the two-way effect using two methods, one is
spontaneous load-assisted induction, and the other is external load-assisted induction
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
17
[17]. However, the shape change obtained is in practice less than that of the one-way
effect.
2.2.3 Pseudo-elasticity
Fig. 2-4 Schematic of stress VS temperature
There is another effect, pseudo-elasticity, also known as ‘super-elastic Effect (SE)’. It
is the shape recovery associated with mechanical loading and unloading of SMAs at
temperatures above AFT and is associated with stress-induced martensitic transformation
and reversal to austenite upon unloading.
The stress dependence of the four transition temperatures can be represented as [18]:
0
T
H
dT
d
r
(2-1)
where is the applied stress; rT is the transformed temperature; H is the
transformation latent heat; T is the temperature and 0 the strain resolved along the
direction of the applied stress.
TMF TMS TAS TAF
Nonlinear
region
Temperature
Str
ess
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
18
As is shown in Fig. 2-4, the temperatures of AST and AFT are highly nonlinear at
low stress levels. Then, the stress dependence of the transformation temperatures can be
expressed as a generalized temperature, given by [19]:
mcdT
d 1
, or 0TcT m (2-2)
where mc/1 is the stress rate, T is the stress dependent transformation temperature and
0T is the zero stress transformation temperature.
2.3 Micro and Macro Analysis of SMA
Concerning about the SMA, it is necessary to introduce changes of crystal structures
and macro shapes during the heating and cooling processes. Many metals have several
different crystal structures at the same composition, but most metals do not show the
shape memory effect. The special property that allows SMA to revert to their original
shape after heating is that their crystal transformation is fully reversible. In most crystal
transformations, the atoms in the structure will travel through the metal by diffusion,
changing the composition locally, even though the metal as a whole is made of the same
atoms. A reversible transformation does not involve this diffusion of atoms, instead all
the atoms shift at the same time to form a new structure, much in the way a
parallelogram can be made out of a square by pushing on two opposing sides. At
different temperatures, different structures are preferred and when the structure is
cooled through the transition temperature, the martensitic structure forms from the
austenitic phase.
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
19
Fig. 2-5 Schematic of micro and macro phenomenon of SMA
The unique property of Ni-Ti alloys is shape memory effect, which can be explained
in a rough 2-dimensional approximation of the underlying atomic rearrangements,
including the micro and macro analysis of SMA. As shown in Fig. 2-5, the atomic
lattice is primarily in the martensite phase when the temperature is less than AST . The
length of SMA wire is LL when it is applied external force; as the temperature
exceeds AST , austenite layers begin to form. Then, austenite phase completes when the
temperature exceedsAFT . The atomic lattice is primarily in the austenite phase. The
length of SMA wire is L , which is the original length without applying external force;
as the alloy cools, martensite layers begin to form when the temperature decreases
below MST and the atomic lattice is primarily in the martensite phase when the
L L
L+ L
Austenite
T AFT
Twinned Martensite
T MFT
Detwinned Martensite
T AST
Cooling
Heating
External deformation
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
20
temperature reachesMFT . The length is practically unchanged and it is the same as in the
austenite phase [20].
2.4 SMA Actuators
SMA is used as actuators because of the following advantages.
SMA actuators offer the highest power-to-weight ratio compared with different types
of actuator technologies [21]. Since An SMA actuator only uses the shape recovery of
the alloy and it can be actuated directly via Joule heating, it does not require any
reduction gear system nor other moving parts, resulting in saving material, production,
maintenance costs and easy miniaturisations of simple actuator systems [22]. In addition,
there is no need for friction mechanisms in SMA actuators, such as reduction gear, it
avoids the production of dust particles, sparks and noise. These properties make SMA
actuators highly attractive for miniature applications. Therefore, it is extremely suitable
for areas, such as microelectronics, biotechnology and medical applications, to apply
SMA as actuators.
SMA have many advantages over traditional actuators, but do suffer from a series of
limitations that may impede practical application [23]. Since SMA actuators are
typically actuated electrically, deactivation typically occurs by free convective heat
transfer to the ambient environment. Consequently, SMA actuation is typically
asymmetric, with a relatively fast actuation time and a slow deactuation time. SMA are
subject to structural fatigue [24-26], which is a failure mode by which cyclic loading
results in the initiation and propagation of a crack. It eventually results in catastrophic
loss of function by fracture. The physics behind this fatigue mode is accumulation of
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
21
microstructural damage during cyclic loading. In addition, SMA are also subject to
functional fatigue, whereby the SMA do not fail structurally, but, due to a combination
of applied stress, and/or temperature, lose the ability to undergo a reversible phase
transformation. At last, SMA actuators are typically actuated electrically by Joule
heating. If the SMA are used in an environment where the ambient temperature is
uncontrolled, unintentional actuation by ambient heating may occur.
SMA spring SMA wire
SMA wire SMA wire
SMA wire Bias spring
SMA spring Bias spring
(1) (2)
(3)
(4)
(5)
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
22
Fig. 2-6 Schematic of different SMA actuators
For SMA actuators, SMA wire or spring is used as actuator in practical application in
order to obtain light and tight manipulators. Because SMA actuators utilize the one-way
effect and can only contract in one direction, it is necessary to provide a biasing force to
pull it back to the original length using a dead weight, a bias spring, or another SMA
element in a differential arrangement. According to [27-28], the primary actuator joint
applications can be divided into ten basic types, as shown in Figure 2-6.
SMA spring SMA spring
Bias spring
SMA spring SMA spring
SMA spring Bias spring
SMA wire
SMA wire
SMA wire
(6)
(7) (8)
(9) (10)
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
23
2.5 Literature Overview
2.5.1 Modeling of SMA
To simulate the behavior of SMAs and as a control design aid, numerous models are
proposed to represent or explain the characteristics of SMAs, most notably in terms of
their thermomechanical relations and the hysteresis effects. Kuribayashi experimentally
observed the relationship between small variations in the force and strain of an SMA
wire, which can be expressed by the following mathematical model [29].
xbuaf (2-3)
where f , u , and x are the force, voltage, and strain, respectively. a and b are
gain constants. Regarding the static model of Eq. (2-3) as the steady state of a dynamic
system, a dynamic model by adding first order terms for G(s) and H (s) in the Laplace
domain is proposed in the following relationship.
)()()()()( sxsbHsusaGsf (2-4)
A thermomechanical law that governs the stress-strain behavior of the SMA element
was proposed by Tanaka, which is expressed by [30-31]
TD (2-5)
where is the Piola-Kirchhoff stress, is the Green strain, T is the temperature,
and is the martensite ratio. D , , and are respectively the elastic modulus,
the thermoelastic, and the transformation tensor.
The phase transformation is described by exponential functions, including heating
and cooling transformation processes. The ratio of martensite of the cooling process
for austenite to martensite can be given by [32]
0])(exp[1 MMSM bTTa (2-6)
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
24
where Ma and
Mb are positive material parameters, 0 is the initial volumetric
fraction when phase transformation takes place. For the reverse transformation, from
martensite to austenite, it can be expressed by
])(exp[0 AASA bTTa (2-7)
where Aa and
Ab are positive material parameters.
Based on the Tanaka’s original model, Boyd and Lagoudas rewrite Tanaka’s original
model, for a three-dimensional theory construction [33]. The constants Ma ,
Mb ,Aa , and
Ab are estimated by
MFMS
MTT
a
)10ln(2
, M
MM
C
ab ,
ASAF
ATT
a
)10ln(2
,A
AA
C
ab (2-8)
Liang and Rogers proposed an alternative equation to represent the martensite
fraction based on cosine function [34]. This model was applied to acoustic vibration
control studies and its results show good agreement with experimental data [35-36]. The
model can be given by
2
1)](cos[
2
1 00
M
MFMC
TTA , Heating process (2-9)
where )()( MFMMSM TTCTTC , MC is a material parameter.
1)](cos[2
0 A
ASAC
TTA
Cooling process (2-10)
where )()( ASAAFA TTCTTC ,AC is a material parameter. And the
MA and AA
can be defined by
MFMS
MTT
A
, ASAF
ATT
A
(2-11)
Since an antagonistic arrangement of SMA actuators was used in Grant’ experiment,
a single linear transformation kinetics equation without hysteresis effect which
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
25
simplifies the model is proposed [37].
AF
AFAS
ASAF
AF
AS
TT
TTTTT
TT
TT
0
1
Heating process (2-12)
Dutta and Ghorbel proposed a differential hysteresis model to represent the SMA
behavior [38]. The operation of the SMA actuator involves different physical
phenomena, such as heat transfer, phase transformation with temperature hysteresis,
stress-strain variations and electrical resistance variation accompanying the phase
transformation. The martensite fraction model of major hysteresis loop is given by
)]2
(1[2
1)()(
/
/''
///
uerfduuguhv u (2-13)
where subscripts + and – denote heating and cooling curves, respectively. / and
/ are constants. g and
g are slops functions which can be given by Duhem
model [39]
0)0(
)))((),(()))((),((
vv
utvtugutvtugv (2-14)
where 2/))(()(
uuu , and g , )( 20 IRCg . )(tu and )(tv are the input and
output.
The slop function of the minor hysteresis loop is given by
)2
)(exp(
2)(
2
/
2
/
/
//
unug i (2-15)
where Ni 1 , /in [0, 1].
The resistance of austenite or martensite phase is given by
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
26
)()21(4
/2
0
0/ T
d
LR MAMA
(2-16)
)(/ TMA , the electrical resistivity of austenite or martensite phase, can by expressed
by
))(exp()( 321 ambA TTpppT (2-17)
9
02
121 )()](1)[()(
i
i
ambiM TTam
mTerfTqqT (2-18)
where 1p ,
2p , 3p ,1q ,
2q ,1m ,
2m , ambT , and ia ( 9,,0 i ) are constant parameters.
Then, the SMA wire electrical resistance R of major hysteresis loop is given by
MA R
v
R
v
R
//11 (2-19)
Concerning about other models to represent the hysteresis effect of smart material,
Preisach modeling of SMA hysteresis is one of the most successful mathematical
models [40-46].
This model can be considered as an operator which integrates infinite weighted
elementary hysteresis operators over a two dimensional region and it can be expressed
by
ddtvtvHtu T )]([)]([)( (2-20)
where )(tv is the input of hysteresis. )(tu is the output of hysteresis. H is an operator
to transform )(tv to )(tu . is the output of an elementary hysteresis operator
subjected to )(tv . is a density function of variables of and to scale outputs
of relays. T is Preisach plan over hysteresis occurs, which is defined as
vvRT :),( 2 (2-21)
where and denote the increasing and decreasing )(tv , respectively.
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
27
],[)( vvtv indicates the hysteresis input domain.
Generally, the inverse model is derived and used to obtain accurate position tracking
results in position control system [47]. The inverse Preisach model can be expressed
according to the following algorithm.
For a given output )(u at a time and a known )( tv and )( tu of a
hysteresis, increase or decrease )(tv by steps of idv , ni ,,1 , until calculated output
of the Preisach operator be mostly close to )(u , until
0)(,)(])1()([)(])([
tuudvntvHandundvtvH (2-22)
or
0)(,)(])1()([)(])([
tuudvntvHandundvtvH (2-23)
Therefore, )(tv can be interpolated as
0)(,])1()([])([
])1()([)()(
0)(,])1()([])([
])1()([)()(
)]([)( 1
tudvntvHndvtvH
dvntvHudvtv
tudvntvHndvtvH
dvntvHudvtv
tuHtv
(2-24)
Then, the inverse model is expressed by
)()]]([[)]([)( 1 tutuHHtvHtu (2-25)
Unlike the Preisach model which inverse is obtained numerically, the
Prandtl-Ishlinskii hysteresis model is analytically invertible and therefore can be easily
implemented in the hysteresis nonlinearity control system, such as position tracking
control [48-49]. The operator, as shown in Fig. 2-7, illustrates the Input-output relation
of the classical Prandtl-Ishlinskii hysteresis operator. Suppose that ],0[ TCm is the
space of the piecewise monotone continuous functions and the input ],0[)( TCtu m is
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
28
monotone on each of the sub-intervals ],[ 1ii tt , where
Tttttt Nii ,,,,0 110 . The output of the classical Prandtl-Ishlinskii
model )(ty can be expressed by
drtuFrpty r
R )]([)()( 0 (2-26)
Fig. 2-7 Play hysteresis operator
where )(rp in an inferable positive density function. r is the positive threshold as
Rrrrrr Nii ,,,,0 110 . ][uFris the classical play hysteresis operator
that is analytically expressed for 1 ii ttt ( 1,,1,0 Ni ) by
))]([),(()]([
)0()0),0(()0]([
irrr
rr
tuFtuftuF
wufuF (2-27)
where ),min(,max),( wruruwufr .
Even though the Prandtl-Ishlinskii model has been applied to characterize hysteresis
u
y
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
29
effects of different smart actuators [50-52], the shortcomings of the model (limitation on
characterizing the behavior of system with output saturation or asymmetric input-output
loops) is eliminated using an alternative generalized play operator [53-54]. As shown in
Fig. 2-8, an increase in input v causes the output of the generalized operator z to
increase along the curve l , while a decrease in input v causes the output of the
generalized operator z to increase along the curver , resulting in an asymmetric loop.
According to Eq. (2-27), the output of the generalized play hysteresis operator is
analytically expressed for 1 ii ttt ( 1,,1,0 Ni ) as
))]([),(()]([
)0()0),0(()0]([
irrr
rr
tuStugtuS
zuguS (2-28)
here ,,)(min(,)(max),( zruruzug llr The output of the generalized
Prandtl-Ishlinskii model, gy , can be expressed by
)(tyg drtuSrp r
R )]([)(0 (2-29)
Fig. 2-8 Generalized play hysteresis operator
y
u
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
30
The inverse model based on the generalized Prandtl-Ishlinskii model is also
developed and used as a feedforward compensator for the purpose of mitigating
hysteresis nonlinearities of smart materials [55-60]. The output of the generalized
Prandtl-Ishlinskii inverse model, inversey , is formulated in discrete form as [61-62]
0,)]([)()0(
1
0,)]([)()0(
1
)(
0
^1
0
^1
^
^
ukuFpkup
ukuFpkup
kyN
jr
jl
N
jr
jl
inverse
j
j
(2-30)
,))(( 1
00
^
i
j
ii
j
i
j
jpp
pp 1,,1,0 Ni (2-31)
),(0
^
ij
j
i
ij rrpr
Nj ,,1,0 (2-32)
2.5.2 Controlling of SMA
In the past, a number of methods had been proposed to control SMA actuators. PID
control method is a linear control system that can be used to control SMA actuators.
Asua, Etxebarria and Garcia-Arribas reported that among the linear controllers, PI
with anti-windup has the best results for position control of SMA actuators by the
experimental results [63]. Da Silva proposed a proportional controller which is applied
in active shape control of a flexible beam [64]. Experimental data shows that, in order to
eliminate the steady state error for tracking the step signal, the overshoot and actuator
saturation is unavoidable. In tuning the gains of PID controller for position control of
SMA actuators, for large values of error the proportional gain of the controller should be
large enough to produce sufficient control effort for error compensation. On the other
hand, for small values of error, a large proportional gain will result in overshoot and
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
31
consequently low performance. Popov et al. proposed a PID controller to control the
SMA actuator while two methods (the Ziegler–Nichols and internal model control
methods) are used to tune the PID controller gains [65]. The simulation results show
that the proposed controller is successful in controlling the displacement of the
transition point position closer to the trailing edge in order to produce a higher laminar
flow region on the airfoil and, therefore, decreases the drag force. Shameli, Alasty and
Salaarieh proposed a PID-P3 controller that adds a proportional cubic term to the
conventional PID controller [66]. They showed that PID-P3 controllers were more
effective than conventional PID controllers for precision position control of a miniature
SMA actuator. Since hysteresis effect of SMA actuator is nonlinear system, it is possible
to develop nonlinear control system to control it. In the literature, it has been shown that
fuzzy logic control is robust in controlling nonlinear systems [67-68]. In the early
1970’s, fuzzy control was first introduced in an attempt to design controllers for systems
that are structurally complicated to model owing to inherent nonlinearities and other
modeling complexities [69]. Cocaud et al. considered fuzzy control of SMA artificial
muscles of a 2 DOF robotic arm [70]. Fuzzy PID controllers have shown good accuracy
and robustness against system nonlinearities and parametric uncertainties. In addition,
PWM controllers are appropriate choices for using as position controllers of SMA
actuators in order to reduce the energy consumption by the actuator. Ma and Song
showed that using pulse width modulation for control of a SMA actuator effectively
saves actuation energy while maintaining the same control accuracy as compared to a
conventional PD controller [71]. They showed that PWM demonstrates robustness to
the external disturbances. Song and Ma proposed an improved PWM technique called
Pulse-Width–Pulse-Frequency (PWPF) modulator which demonstrates that the PD
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
32
controller with PWPF modulation consumes 50% less energy than the one without
modulation [72]. The biomimetic control of an anthropomorphic artificial finger is
presented, which is actuated by three antagonistic SMA muscle pairs that are each
configured in a dual spring-biased configuration and it focuses on the design and
experimental verification of a new fuzzy pulse-width-modulated
proportional-integral-derivative (i.e. fuzzy PWM-PID) controller that is capable of
realizing cocontraction of the SMA muscle pairs, as well as online tuning of the PID
gains to deal with system nonlinearities and parameter uncertainties [73]. To maintain a
desired position of a joint, the corresponding agonistic muscle pairs are cocontracted by
the central nervous system and stiffen the joint. Both numerical and experimental results
show the performance advantage of the cocontracting fuzzy PWM-PID controller over
the original PWM-PID controller.
Furthermore, many researches had worked on nonlinear control aspects of SMA
actuators to solve the hysteresis problem in the past. Grant and Hayward demonstrated a
variable structure controller which is applied to a pair of antagonist actuators [74]. In
this thesis, the feedback switches between the two actuators according to the sign of the
displacement error. In addition, a further improvement was added to compensate for
known gross nonlinearities by modulating the current magnitude in a discrete manner as
a function of the state space position. Therefore, it is possible to realize smooth and
robust control with very little cost in complexity. Due to modeling uncertainty,
nonlinear behavior of the system and classic control methods such as
Proportional-Integral-Derivative control are not able to present fast and accurate
performance, a nonlinear robust control algorithm for accurate positioning of a single
degree of freedom rotary manipulator actuated by SMA based on Variable Structure
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
33
Control is presented [31]. The model includes nonlinear dynamics of the manipulator, a
constitutive model of SMA, and electrical and heat transfer behavior of SMA wire.
Computer simulation and experimental results for different stabilization and tracking
situations are presented and results show fast, accurate, and robust performance of the
control system. A gain-scheduled controller for an SMA actuator is presented [75]. A
model has been proposed based on concepts from physics in order to achieve accurate
control of an SMA actuator which includes Joules heating-convectional cooling to
explain the dynamics of temperature, Fermi-Dirac statistics to explain the variation of
mole fraction with temperature, and a stress-strain constitutive equation to relate
changes in mole fraction and temperature to changes in stress and strain of the SMA.
Then, this model is applied to develop a gain-scheduled controller to control the strain
in the SMA. Simulation and experimental results show fast and accurate control of the
strain in the SMA actuator.
In the second approach of nonlinear controllers, there are many researches focused on
the inverse model control system to compensate the hysteresis effect of SMA nonlinear
system because of its effectiveness and flexibility. As mentioned above, the inverse
models based on the inverse Preisach model [47] or the generalized Prandtl-Ishlinskii
model [55-60] are also developed and used as a feedforward compensator of smart
materials. Neural networks, which possess properties of nonlinear function mapping and
self-adaptation, have been used to model hysteresis and, in some cases, to compensate
for hysteresis in SMA actuators [76]. In addition, a neural network inverse model and a
sliding-mode based robust feedback controller are used to compensate the SMA
hysteresis phenomenon [77]. Since the inverse model was not exact, three more control
signals (PD controller, feedforward controller, and sliding-mode base robust
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
34
compensator) were modified to improve the control performance in both open-and
closed-loop fashions. Moreover, it would not be exact if there was a change of load or
working environment because the inverse model was trained offline. Webb and
Lagoudas proposed an adaptive hysteresis model for on-line identification and
closed-loop compensation because of inaccurate model based on off-line identification
[78]. Hysteresis compensation based on the Krasnosel’ skii-Pokrovskii hysteresis
inverse model is also presented [79]. To alleviate the problem of direct temperature
measurements of the SMA wire, an observer based on a simplified thermal model of the
SMA wire that requires only rough estimates of the thermal parameters is implemented.
The reference input must be sufficiently rich in order to update the parameters of the
Krasnosel’ skii-Pokrovskii hysteresis model and therefore cannot be used for step
regulation tracking.
2.6 Chapter Summary
This chapter provides essential background information on shape memory alloys,
including the concept of phase transformation temperature, types of SMA, micro and
macro analysis of SMA, modeling and control methods.
Since SMA have many advantages, it is used in commercial or industrial actuator
applications. However, there remains obstacles in developing SMA actuators, such as
slow speed and the difficulty of accurate position tracking control, as well as energy
inefficiency. In order to model and control SMA actuators, many researches had been
developed in the past. And they are introduced and summarized in this chapter.
This dissertation aims to provide some groundwork on practical control strategies to
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
35
achieve rapid response speed of SMA actuators. Results and significant work that have
been accumulated during this Ph.D. research will be documented and described in depth
in the following chapters. It is hoped that this thesis will be useful for further research of
SMA, including modeling and control of SMA actuators.
CHAPTER 2: NICKEL-TITANIUM SHAPE MEMORY ALLOYS
36
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
37
CHAPTER 3:
LATENCY DURATION INVESTIGATION OF
SMA ACTUATOR
3.1 Introduction
There have been some discussions over the years as to whether shape memory alloys
can respond very quickly. As noted in the literature review of chapter 2, researchers
have attempted to improve upon the controllable speed of SMA actuators. Some of the
results are quite significant, especially in the small or micro-actuator scale [31].
In this chapter, we will investigate the possibility of SMA actuators having very rapid
and detectable responses when subjected heating and cooling. The motivations for this
investigation of SMA will first be explained. In section 3.3 and 3.4, the method and
experimental setup will be described, respectively. The results of the experiments are
presented in section 3.5. Some discussions and conclusions about the results are also
presented in section 3.6 and 3.7, respectively.
3.2 Motivation and Target
In order to reduce the latency duration, it needs to be investigated. The objectives of
the experiments described in this chapter are to determine, firstly, there is hysteresis
effect for SMA wire and the cycle time of different heating time with binary control is
different, and secondly, if the latency duration exists due to the hysteresis effect. As
shown in Fig. 3-1, the output displacement increases from zero to S during heating
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
38
processes ac when the temperature increases, and decreases from S to zero during
cooling process cf when the temperature decreases. However, the output displacement is
unchangeable during the heating process ab and cooling process cd even though the
temperature of SMA increases and decreases, respectively. The time Tab (from a to b)
and Tcd (from c to d) is latency duration of SMA actuator. In this thesis, only the latency
duration Tcd caused by the nonlinear effect of SMA is investigated. It is useful to find
out if the latency duration in an SMA wire can be produced in order to obtain maximum
output displacement with binary control. The results of this investigation will be
important for SMA control in the following chapters.
Fig. 3-1 Schematic of output displacement for SMA wire
3.3 Method
Hysteresis and significant nonlinearities in the behavior of SMA actuators encumber
effective utilization of these actuators, especially actuators using thick SMA wires
which have long latency duration. Instead of controlling temperatures with a Peltier
a b c f d
Dis
pla
cem
ent S
0
Heating
Cooling
Tcd
Time
Tab
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
39
device, this thesis uses electricity to heat the SMA wires with ‘on-off’ binary control [4].
As shown in Fig. 3-2, Joule resistive heating causes the SMA actuator to output
displacement which increases from zero during the heating process ac when power is
turned on. The output displacement is steady during ab which is caused by the
hysteresis effect in heating process. When power is turned off, the SMA actuator is
cooled with natural convection and the output displacement reaches zero at f during the
cooling process cf (shown with black line1). Since the output displacement is
unchangeable during cd, the latency duration is Tcd.
Fig. 3-2 Schematic of the binary control with latency duration Tcd
The latency duration caused by hysteresis (Tab and Tcd in Fig.3-2) slows both the
heating and cooling response speed, especially during the phase transformation of a
Time
a b c
S
Dis
pla
cem
ent
Volt
age
0
f d
0
On
Heating
Cooling
Off
Tcd
1
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
40
thick SMA wire from austenite to martensite. This greatly limits the response of SMA
actuators. The heating time is much shorter than the cooling time when the SMA wire is
heated by a large heating current.
3.4 Experimental Setup
Fig. 3-3 Schematic of experimental setup
In order to obtain the output displacement of SMA wire, the experimental setup is
built shown in Fig. 3-3. The fixed end of the wire is connected to a load cell and the
other end is attached to a bias spring. When the SMA wire is heated to achieve the
austenite length and the electric current is discontinued, the bias spring pulls the SMA
wire back to the martensite length. A reflector (reflecting the laser beam) is connected to
the spring for displacement measurements. As the SMA wire contracts or extends, the
reflector moves forward and backward. Both the data for displacement and force is
received by microcontroller, which will be sent to computer. The controller will
calculate the output voltage to microcontroller which will transform it to PWM to heat
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
41
the SMA wire. 5V is used as the maximum heating voltage to heat the SMA wire. Power
resistor is used to protect the microcontroller.
For the experiment, thick SMA wire is selected. As shown in Fig. 3-4(i), two same
SMA wires (SMA1 and SMA2, 0.5mm in diameter, 140mm in length) are used in the
following tests to obtain the output displacement suffering to different heating time. The
latency duration of both SMA wires will be tested separately. Fig. 3-4(ii) shows the
length of single SMA wire. As shown in Fig. 3-5, three output voltages are measured for
each SMA wire to obtain the variation of resistance. In order to make sure the maximum
input current is about 2A which can heat the SMA wire to finish the austenite phase,
small resistor (1.1Ω) is added in the driver circuit. The resistance calculation function is
denoted by
)/()( 3221 vvvvRsma (3-1)
where 1v ,
2v , and3v , are the input voltages which will be used in chapter 4 and chapter
5. More detailed information is listed in Table 3-1.
Fig. 3-4 SMA wires used in the experiment
140mm
SMA1 SMA2
(i) (ii)
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
42
Fig. 3-6 shows the schematic of the experimental setup to test the latency duration
with different heating time for SMA wire. In these tests, two SMA wires with the same
length are tested. In the following chapters, they will be connected together to verify the
proposed method. As shown in Fig. 3-6(i), it is the photo of experimental setup of SMA
actuator outlined in Fig. 3-2; Fig. 3-6(ii) is the photo of experimental setup; the
displacement and force are obtained by a KEYENCE LC-2000 laser displacement meter
and a TEDEA-HUNTLEIGH load cell, respectively.
Fig. 3-5 SMA wires used in the experiment
Table 3-1 Parameters for the experiments
Ambient temperature 24°C SMA diameter 0.5mm
MOSFET K2232 SMA length 140mm
Power supply 5V Spring stiffness 653.3N/m
Microcontroller Arduino Pretension force 3.27N
Power resistor
SMA SMA
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
43
Fig. 3-6 (i) Photo of the SMA actuator and (ii) Photo of experimental setup
Load cell
Reflector
SMA1 or SMA2
Laser sensor
Bias spring
(i)
(ii)
Laser sensor
SMA structure
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
44
3.5 Results
3.5.1 Results with Different Ambient Temperature
According to the method mentioned in section 3.3, as shown in Fig. 3-7, ‘on-off’
binary control with 5V is applied to heat the SMA. The SMA is heated from t1=5s to
t2=15s. Fig. 3-8 shows the output displacement with input voltage in Fig. 3-7. It shows
that the output displacement is practically unchangeable from a =5s to b=7.5s. From b
to c1=10s, the output displacement increases from zero to the maximum output
displacement 0.81mm during heating process; from c1 to c=15s, the output
displacement is practically unchangeable even heated which means the SMA1 is
overheated; from c to d=35s, the maximum output displacement is 0.9mm when the
power is turned off; the SMA starts the transformation from austenite to martensite
when output displacement is less than 0.81mm from d. Then, the time Tab =2.5s (from a
to b) and Tcd =20s (from c to d) is latency duration of SMA actuator.
The martensite start displacement (0.81mm) is 90 percent of the maximum output
displacement. When there is no overheating process if the heating time is short, the
output displacement will decline to zero during cooling process, and the 90 percent of
the maximum output displacement can be used as criteria to decide the martensite start
displacement.
Since the temperature of SMA wire is the balance of Joule heat and the heat
dissipation to the ambient environment. Therefore, the ambient temperature will affect
the heating and cooling time. Fig. 3-9 shows the results of different ambient temperature
with input voltage in Fig. 3-7. It clearly shows that the SMA has the similar heating
trajectory with different ambient temperature during the heating process from t1=5s to
t2=15s. With the different ambient temperature, the SMA almost finishes the phase
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
45
transformation from martensite to austenite at P=0.81mm during the heating process at
10s; the SMA obtains the same maximum output displacement at Q =0.90mm during
the cooling process as well.
However, compared with other ambient temperature, the SMA extends the fastest
when the ambient temperature is 18°C during the cooling process and starts
transformation from austenite to martensite at d1=31s. More detailed information about
latency duration with different ambient temperature is listed in Table 3-2.
Fig. 3-7 10s as heating time for SMA
Table 3-2 Latency duration with different ambient temperature
Ambient temperature (°C) 24 21 18
Latency duration (s) 20 19 16
0 20 40 600
1
2
3
4
5
6
Time (s)
Volt
age
(V)
t1 t2
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
46
Fig. 3-8 Output displacement for SMA
Fig. 3-9 Results with different ambient temperature for SMA
0 20 40 60-0.5
0
0.5
1
1.5
0 20 40 60-0.5
0
0.5
1
1.5
21°C
18°C
24°C
P
Time (s)
d1 d
Dis
pla
cem
ent
(mm
)
a c
a c b d c1
Cooling Heating
Time (s)
0.9mm
0.81mm
Q
24°C
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
47
3.5.2 Results with Binary Control
Here, the latency duration and cycle time with different heating time will be
investigated for both SMA wires in ambient temperature 24°C in the following
experiments. The latency duration of both SMA1 and SMA2 is tested. Fig. 3-10 and Fig.
3-11 show the results of different heating time and output displacement of SMA1,
respectively. It is obvious that different heating time leads to different latency duration.
As shown in Fig. 3-11, when the heating time is 4s, the maximum output displacement
is small, about 0.75mm, and the SMA1 starts transformation from austenite to
martensite at d1=16.6s; when the heating time is larger than 5s, such as 7s, the SMA1
will be overheated, and the SMA1 starts transformation from austenite to martensite at
d2=32s. More detailed information about latency duration is listed in Table 3-3. As
mentioned in section 3.3, the SMA finishes martensite phase when the output
displacement reaches zero. Since there are small offset from the position zero when
SMA1 finishes the phase transformation from austenite to martensite during the cooling
process, a horizontal line along the martensite completion part is used to decide the
martensite finish time. According to this criteria, the SMA1 finishes the martensite
phase at f=55s when the heating time is 4s.
Table 3-3 Latency duration with different heating time for SMA1
Heating time (s) 10 9 8 7 6 5 4
Latency duration (s) 20 18.6 18. 5 17 16 13 7.6
Maximum
Displacement S(mm)
0.9
0.9
0.9
0.9
0.9
0.84 0.75
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
48
Fig. 3- 10 Heating time of SMA1
Fig. 3- 11 Output displacement of SMA1
in ambient temperature 24°C
0 20 40 600
1
2
3
4
5
6
10 s9 s8 s7 s6 s5 s4 s
0 20 40 60-0.5
-0.3
-0.1
0.1
0.3
0.5
0.7
0.9
1.1
1.3
1.5
10 s9 s8 s7 s6 s5 s4 s
Time (s)
Time (s)
Dis
pla
cem
ent
(mm
)
d1
Offset
Martensite completion
Horizontal line
f d2
Volt
age
(V)
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
49
In order to prevent from overheating exactly, the heating time 5s is used to obtain
maximum output displacement which is 0.84mm; As shown in Fig. 3-12, the SMA1
starts the phase transformation from austenite to martensite at d =23s when the
martensite start displacement (0.76mm) is 90 percent of the maximum output (0.84mm).
Then the SMA1 finishes the cooling process at f=60s. Fig. 3-13 shows the results of
output displacement with different heating time for SMA2. It is obvious that different
heating time leads to different output displacement as well. Since the maximum output
displacement (0.71mm) during the heating process is almost the same as that during the
cooling process when SMA2 is overheated, the martensite start displacement can not be
decided by percentage of maximum output displacement. Therefore, the latency
duration of SMA2 with different heating time, which is overheated or not, is decided by
the same criteria (90%) as SMA1.
Fig. 3-12 Binary control with latency duration Tcd for SMA1
in ambient temperature 24°C
0 20 40 60-0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
a b c
f
d
Tcd
Time (s)
Dis
pla
cem
ent
(mm
)
0.84mm 0.76mm
S
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
50
Fig. 3- 13 Output displacement of SMA2
in ambient temperature 24°C
Fig. 3-14 Binary control with latency duration Tcd for SMA2
in ambient temperature 24°C
0 20 40 60-0.5
0
0.5
1
1.5
0 20 40 60-0.5
0
0.5
1
1.5
5 s4s3.5 s3 s2.6 s2.3 s
a b c
f
d
Tcd
Time (s)
Time (s)
Dis
pla
cem
ent
(mm
) D
ispla
cem
ent
(mm
)
0.64 mm
0.71mm
S
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
51
Table 3-4 Latency duration with different heating time for SMA2
Heating time (s) 5 4 3.5 3 2.6 2.3
Latency duration (s) 23 15.5 16 14 10.7 9.9
Maximum
Displacement S(mm)
0.71
0.71
0.71
0.71
0.68 0.64
Fig. 3-13 also shows that when the heating time is 5s, the SMA wire will be
overheated which leads to take longer time to decline when the power is turned off. In
order to prevent overheating exactly, the heating time is set 3s to obtain maximum
output displacement S= 0.71mm for SMA2; when the heating time is 2.3s, the output
displacement is about 0.64mm. More detailed information is listed in Table 3-4.
Fig. 3-14 shows results of the latency duration with heating time 3s for SMA2. The
SMA2 completes the phase transformation from austenite martensite during heating
process ac. And SMA2 completes the phase transformation from martensite to austenite
during cooling process cf. According to the criteria mentioned above, the martensite
start displacement is 0.64mm. The SMA2 starts the martensite phase at d=22s and
finishes the cooling process at f=59s. Comparison about the latency duration
corresponding to Fig. 3-1 is listed in Table 3-5.
Table 3-5 Heating time and cycle time of SMA wires
Tab Tbc Tcd Tdf Cycle time
SMA1 1.8s 3.2s 13s 36s 54s
SMA2 1.7s 1.3s 14s 37s 55s
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
52
3.6 Consideration and Discussion
The prototype system developed for the experimental investigation of latency
duration has demonstrated the feasibility. There are, however, a few critical points
needing further consideration.
1. As shown in Fig. 3-8, the output displacement slightly changes from a to b and
the minimum output displacement is -0.08mm. In addition, the output
displacement slightly decreases from c1 to c when heated. The reason is when the
temperature of SMA increases, the SMA will expand due to the function that
metals expand when heated and contract when cooled. When the temperature of
SMA is larger than austenite start temperature from b, and when the temperature
of SMA is lower than austenite finish temperature from b to c1, there will be
output displacement caused by the shape memory effect. Compared with the
output displacement caused by shape memory effect, the displacement caused by
thermal expansion can be ignored during bc1. Moreover, the output displacement
slightly increases and the maximum output displacement is 0.9mm during cooling
process cd. The reason is when the temperature of SMA decreases, the SMA will
contract due to the function that metals expand when heated and contract when
cooled. When the temperature of SMA is lower than martensite start temperature
from d, there will be output displacement caused by the shape memory effect and
the displacement caused by thermal contraction can be ignored.
2. As shown in Fig. 3-9, the output displacement is almost the same with different
ambient temperature during the heating process from 5s to 15s. Because heating
voltage is large enough to make the SMA wire finish the phase transformation
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
53
from martensite to austenite for each test. And the heat loss with different ambient
environment can be ignored during heating process. However, the output
displacement decreases the fastest when the ambient temperature is 18°C, because
the heat loss in 18°C from the SMA to ambient environment is faster than in 22°C
and 24°C. As shown in Table 3-2, the difference of latency duration and cooling
time for different ambient temperature is caused by different heat loss rate in air
[38]. Therefore, in order to obtain correct latency duration, the ambient
temperature should be steady.
3. As shown in Fig. 3-11 and Table 3-5, the latency duration with heating time 10s
(overheated) is 20s which is larger than that with heating time 5s (13s). It takes
much longer time to decline during the cooling process once the SMA is
overheated. Therefore, in order to obtain rapid response speed for SMA actuator,
it is important to control the heating time to avoid overheating.
4. As shown in Fig. 3-12 and Table 3-5, since the heating voltage is large, the
latency duration Tab during the heating process is much smaller compared with
latency duration Tcd. If the heating voltage is small, the latency duration Tab will
be more obviously. As shown in Fig. 3-15, latency duration Tab for input voltage
2.94 V is 4.2s which is larger than with input voltage 5V (1.8s) because the
temperature of SMA increases slower. And, there will be no output displacement
when the input voltage is small, such as 1V because the temperature is less than
martensite start temperature. It is possible to shorten the latency duration Tab with
more segments, such as 4 SMA wires connected together.
5. As shown in Fig.3-12, Fig.3-14, and Table 3-5, due to the specimen error in
length and material, there are small differences about the maximum output
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
54
displacement (0.84mm for SMA1 and 0.71mm for SMA2) and latency duration
(13s for SMA1 and 14s for SMA2) even though the SMA wires are tested in same
ambient temperature.
6. Fig. 3-16 shows the results with thin SMA wire (0.15mm in diameter) and thick
SMA wire (0.5mm in diameter) with the same maximum output displacement
(0.56mm) and 140mm in length with ambient temperature 21°C, the SMA
finishes the heating process at c and starts the martensite phase at d. Then the
latency durations for both SMA wires are 0.5s and 2s. The output displacement of
both wires reaches zero at 20s and 55s during the cooling process when the power
is turned off, respectively. It clearly shows that the thick SMA wire has longer
latency duration and slower cooling response speed, because it has slower heat
loss ratio than thin one.
Fig. 3-15 Binary control with different heating voltage for SMA1
in ambient temperature 24°C
0 20 40 60-0.5
0
0.5
1
1.5
a b Time (s)
Heating Cooling
5V
1V
2.94V
Dis
pla
cem
ent
(mm
)
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
55
Fig. 3-16 (i) SMA wires; (ii) Results for SMA 0.15mm in diameter; (iii) Results for
SMA 0.5mm in diameter in ambient temperature 21°C
0 20 40 600
1
2
0 20 40 60
0
0.2
0.4
0.6
0 20 40 600
2
4
6
0 20 40 60
0
0.2
0.4
0.6
Dis
pla
cem
ent
(mm
) V
olt
age
(V)
Time (s)
Dis
pla
cem
ent
(mm
) V
olt
age
(V)
Time (s)
c
d
c d
(iii)
(ii)
(i)
0.5mm 0.15mm
CHAPTER 3: LATENCY DURATION INVESTIGATION OF SMA ACTUATOR
56
3.7 Chapter Summary
The experimental results show that there is hysteresis effect in SMA wire. Due to the
hysteresis effect, it is reasonable to test the latency duration caused by the hysteresis
effect with the binary control.
It has been demonstrated in this chapter that the designed experimental setup can
meet the need of demand to detect the displacement and a 0.5mm diameter NiTi SMA
wire have rapid and detectable responses during the heating process when applied 5V
compared with the cooling process.
Experiments not only indicate the existence of the latency duration at room
temperature which is arisen from the hysteresis effect of SMA wire, but also the
variation of it with different heating time. As shown in Table 3-5, with the proper
heating time to obtain the maximum output displacement and eliminate the overheating,
the latency duration is about 24.1% and 25.4% of cycle time for SMA1 and SMA2,
respectively. We have observed there is small difference between two SMA wires such
as the maximum output displacement.
These observed results are very important to the work in the following chapters.
Based on the results, it is proposed that some new control method may be capable of
improving the speed of response by shortening the latency duration during the cooling
process.
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
57
CHAPTER 4:
PHASE RESISTANCE FEEDBACK CONTROL
TO ACHIEVE RAPID RESPONSE SPEED
4.1 Introduction
Although SMA appears to be attractive for robot applications, they also come with
several limitations. First, since their operating principle is based on a phase transition in
a metal, they have highly nonlinear properties such as the stress-strain relationship,
internal resistance, latent heat of transformation, and thermal conductivity are all phase
dependent. Second, their input-output relations contain a wide hysteresis loop, making
them difficult to control accurately. The third major limitation of SMA is efficiency. The
energy efficiency of SMA is theoretically restricted to approximately 10%. Efficiency is
often less than 1% in practical applications, since the driving principle of the actuator
can be considered as a heat engine operating at low temperatures. Hence, applications of
SMA actuators must be directed at areas where energy efficiency is not a concern. SMA
also have two major inherent mechanical limitations: limited percent strain and low
bandwidth. The absolute percent strain is approximately 8%. With practical applications
restricted to around 5%. The motion bandwidth of SMA actuator is generally low.
Many research had focused on improving the response speed of the SMA, including
antagonistic-pair arrangements [80], variable structures [74], and two-stage relay
control [37]. Thin SMA wires, thinner than 0.2mm, can achieve the aim as actuators
because the cooling speed is fast in air. Thick SMA wires, thicker than 0.5mm, have
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
58
long latency durations, making it difficult to obtain rapid response speeds. Some
researchers have shown that active cooling chips [4, 81] and variable structure control
can improve reaction times of SMA actuators [74] (Fig. 4-1). As a function of
temperature, SMA changes the shape through a metallographic transformation that
results in resistance changes. The relationship between electrical resistance and
displacement of the SMA has been investigated [13]. Resistance also can be used as
feedback in the control system of SMA actuator. SMA actuators are chosen as an
example of a plant with hysteresis and a control system to compensate the hysteresis is
proposed controlling the electrical resistance of SMA, which reflects its state, and the
usefulness of the controller is confirmed experimentally [82].
Fig. 4-1 An illustration of variable structure control [74]
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
59
In chapter 3, the latency duration of SMA wire is investigated. In this chapter, we will
investigate the possibility of SMA actuators having rapid response when using phase
resistance as feedback to shorten the latency duration. The motivations will first be
explained. In section 4.3 and 4.4, the method and results will be described, respectively.
Some discussions and conclusions about the results are also presented in section 4.5 and
4.6, respectively.
4.2 Motivation and Target
Fig. 4-2 Schematic of the binary control with latency duration Tcd
To obtain rapid response speed for SMA actuator, an alternative approach, PRFC
(Phase resistance feedback control), is proposed to control the phase resistance in a
closed-loop feedback system of two connected thick SMA wires. There are two basic
Time a b c
S
Dis
pla
cem
ent
Volt
age
0
f d
0
On
Heating
Cooling
Off
Tcd
e
1
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
60
concepts underlying the PRFC. One is the introduction of phase resistances which
divide the hysteresis loop into four parts. The other is to utilize the phase resistances as
the feedback to reduce the latency duration. Experiments to prove the effectiveness
demonstrated that the latency duration caused by the hysteresis effect was shortened and
showed the advantages of the proposed method. Experimental results that demonstrate
the advantages and justify the concepts are also presented. As discussed in chapter 3, the
traditional control method for position is an ‘on-off’ binary control. As shown in Fig.
4-2, Tcd is the latency duration which has been investigated in chapter 3. Here, PRFC is
proposed to achieve rapid response speeds, like the dotted line 1 (Fig. 4-2). The Joule
resistive heating causes the SMA actuator to contract during the heating process ac
when power is turned on. The SMA actuator is cooled with natural convection which
extends from c and completes the cooling process at e, shortening the latency time Tcd.
4.3 Method
4.3.1 Phase Resistance
The SMA may be seen as a smart material that changes shape due to changes in
temperature. These changes are reversible. The shape memory effect arises from
temperature and stress dependent shifts in the crystalline structure of the SMA, shifts
between martensite and austenite phases. During the phase transformation processes, the
resistance changes with changes in temperature. As with the strain-to-temperature or
resistance-to-temperature relationship, the strain-to-resistance curve exhibits a
hysteresis loop [83-85]. As shown in Fig. 4-3 and Table 4-1, there are four phase
transformation resistances Ras, Raf, Rms, and Rmf. In the heating process, from A to B, the
resistance increases from Rmf to Ras without changes in the output strain; from B to C,
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
61
the resistance decreases from Ras to Raf as the SMA transforms from martensite to
austenite. In the cooling process, from C to D, the resistance decreases from Raf to Rms
without changes in the output strain, while from D to A, the resistance increases from
Rms to Rmf as the SMA transforms from austenite to martensite.
Table 4-1 List of important data in Fig. 4-3
Ras Austenite start resistance
Raf Austenite finish resistance
Rms Martensite start resistance
Rmf Martensite finish resistance
Tab Heating time from A to B
Tbc Heating time from B to C
Tcd Cooling time from C to D
Tdf Cooling time from D to A
Fig. 4-3 Phase transformation resistances VS strain
Mar
tensi
te
Tab
Au
sten
ite
0
100%
Rms Raf Rmf Ras
B A
Tcd
C D
Tbc
Heating
Tdf
Cooling
Resistance
100%
0
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
62
4.3.2 Phase Resistance Feedback Control Method (PRFC)
As shown in Fig. 4-3, we have divided the heating and cooling process into four parts
with times Tab, Tbc, Tcd, and Tdf, as listed in Table 4-1, and a cycle time T is defined as
T=Tab+Tbc+Tcd+Tdf (4-1)
The SMA shrinks during the time Tbc rather than Tab in the heating process and
extends during the time Tdf rather than Tcd in the cooling process. If the gap between the
heating and cooling lines (Fig. 4-3) is large, then the resistance-phase state
characteristics of SMA have strong hysteresis properties. In this case it is simple to
identify phase resistances.
The critical aspect of the PRFC method is to shorten the total cooling time by
controlling the phase resistances with two connected SMA wires (SMA1 and SMA2 in
Fig. 3-4) that are coordinated. Fig. 4-4 shows the basic theory behind this method, using
the definitions in Table 4-1, and may be explained as follows.
As shown in Fig. 4-4(i), in section AC, the resistance of SMA1 is maintained at Rms,
and there is no displacement output from the SMA1 wire. The resistance of SMA2
increases from Rmf to Ras and then decreases to Raf. The SMA2 wire completes the phase
transformation from martensite to austenite. As shown in Fig. 4-4(ii), the SMA2
elongates and then contracts, which respectively leads to positive and negative
displacements during the heating process AC; the maximum output displacement is at S.
The heating time Th can be described by
Th=TAB+TBC=Tab+Tbc (4-2)
In section CD, once the resistance of SMA2 reaches Raf, the SMA1 is rapidly cooled
without any current and the resistance of SMA2 is maintained at Rms. The result is that
there is no displacement output from the SMA2 wire. The resistance of SMA1 increases
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
63
to Rmf and the SMA1 wire completes the phase transformation from austenite to
martensite. During the cooling process, CD, the displacement of SMA1 is positive. The
cooling time Tc can be given as follows
Tc =TCD=Tdf (4-3)
In addition, the time of a cycle for one SMA wire Tsingle is longer than with the binary
control method. It is expressed as follows
Tsingle=Tab+Tbc+TCE+Tdf (4-4)
where TCE > Tcd
Combining the two SMA wires, the latency time Tcd can be eliminated. Therefore the
time of a cycle in PRFC, Tp, is given as
Tp=Th+Tc= Tab+Tbc+Tdf (4-5)
Fig. 4-4 Schematic for the PRFC method
B D E C A Time
Raf
Rms
Ras Rmf
Dis
pla
cem
ent
SMA2
SMA1
Res
ista
nce
0
0
Tab
S
(i)
(ii)
Tbc Tdf Tbc Tab
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
64
Fig. 4-5 Output displacement with the PRFC
To explain the PRFC method in more detail, the output displacement of SMA actuator
is demonstrated. As shown in Fig. 4-5, the displacement is zero when both SMA wires
are in the martensite phase before the PRFC.
Step 1: at the beginning of PRFC, SMA1 is in the martensite starting phase and
SMA2 is in the martensite finishing phase, the output displacement is S.
Step 2: SMA1 is in the martensite starting phase and SMA2 completes the
transformation from martensite to austenite, the output displacement is S as well.
Step 3: once the SMA2 completes the transformation from martensite to austenite,
SMA1 starts the transformation from austenite to martensite. At the same time, SMA2
remains at martensite starting phase, the output displacement is S; Combining the two
SMA wires, Tcd is shortened and the maximum output displacement from step 1 to step
3 is S.
4.3.3 PID controller
The phase resistance Rms needs to be used as feedback to maintain the output
displacement precisely by tuning the PID parameters. A proportional-integral-derivative
controller (PID controller) is a generic control loop feedback mechanism (controller)
widely used in industrial control systems. A PID controller calculates an "error" value as
SMA1 SMA2 Martensite
finish phase
Step 1
Step 2
Step 3
S
+ - + -
S
0 0
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
65
the difference between a measured process variable and a desired setpoint. The
controller attempts to minimize the error by adjusting the process control inputs. The
PID controller calculation algorithm involves three separate constant parameters, and is
accordingly sometimes called three-term control: the proportional, the integral and
derivative values, denoted P, I, and D, these values can be interpreted in terms of time. P
depends on the present error, I on the accumulation of past errors, and D is a prediction
of future errors, based on current rate of change. The weighted sum of these three
actions is used to adjust the process via a control element such as the position of a
control valve, a damper, or the power supplied to a heating element. By tuning the three
parameters in the PID controller algorithm, the controller can provide control action
designed for specific process requirements. The response of the controller can be
described in terms of the responsiveness of the controller to an error, the degree to
which the controller overshoots the setpoint, and the degree of system oscillation. Note
that the use of the PID algorithm for control does not guarantee optimal control of the
system or system stability. To be able to use the phase resistances as feedback, a PID
controller was installed instead of the traditional ‘on-off’ binary control, and this is
shown in Fig. 4-6. The parameters of the PID controller are based on the value of the
error e. The PWM value is then converted to voltage and heats. Since the
microcontroller can only receive the value from 0 to 255, then the output of PID
controller is converted to this range. The output voltage is from 0 to 5V, and the
following equations represent the error output voltage relationship.
]0,255[,)(
)()( 0 Piddt
tdeKdtteKteKPid d
t
ip (4-6)
where e(t) is the signal error of the resistance.
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
66
255/*5 PidVoutput (4-7)
There are two purposes of using resistance as feedback.
1. It can maintain the output displacement without using displacement as feedback.
2. The SMA wire is ready to cool down quickly without duration during the cooling
process since the latency duration is shortened.
Fig. 4-6 Block diagram of PRFC
4.3.4 Experimental Setup
An experimental setup of the proposed double SMA actuator was made, and as
shown in Fig. 4-7(i), SMA wires are connected by an insulation joint to prevent short
circuiting; the fixed end of the wire at the joint is connected to a load cell and the other
is attached to a bias spring. A reflector is connected to the spring for displacement
measurements. As the SMA wire shrinks and extends, the reflection sheet moves
forward and backward. Fig. 4-7(ii) is the experimental setup explained in Fig. 4-7(i).
Detailed parameters of the experiment are shown in Table 3-1.
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
67
(i)
(ii)
Fig. 4-7(i) Schematic outline of the experimental setup and (ii) Photo of the
experimental setup
Reflector
Laser sensor
Load cell
Power
SMA1 SMA2
Bias spring
Insulation joint
Load cell Insulation joint
Reflector
SMA1 SMA
2
Laser sensor
Bias spring
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
68
Fig. 4-8 Driver circuit
Fig. 4-9 Control loop of the experiment
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
69
Fig. 4-10 Results with different cut-off frequency for SMA1 and SMA2
in ambient temperature 24°C
In order to measure the resistance of both SMA wires, the driver circuit is designed
shown in Fig. 4-8. Two SMA wires are measured and the calculation equation can be
expressed by Eq. (3-1) in chapter 3. The displacement is detected as the same in chapter
3 by microcontroller. As shown in Fig. 4-9, a computer is connected to a microcontroller
with a USB port. The microcontroller sends data used as the feedback to the computer
and receives the control values sent by the computer. Then, the microcontroller sends a
PWM value to the driver circuit to control the input voltage of the SMA wires. Before
identify the phase resistance, we need to filter the data obtained from microcontroller.
The sampling time is 0.05s, and to filter out the high-frequency noise component, a
third order Butterworth filter with a cut-off frequency. Fig. 4-10 shows the results with
different cut-off frequency and Wc=2Hz is used in the experiment.
0 20 40 60 80 1001.4
1.45
1.5
1.55
1.6
Without filterWc=5 HzWc=2 Hz
Time (s)
Res
ista
nce
(Ω
)
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
70
4.4 Results
4.4.1 Phase Resistance Identification
To identify phase transformation resistances, it is necessary to determine the major
hysteresis loop of the displacement-to-resistance of SMA wires involved, and it is also
important to determine the heating time, especially to prevent overheating.
In chapter 3, the heating time 5s is used to heat the SMA wire to obtain maximum
output displacement instead of overheating. Fig. 4-11 shows the results of resistance
with heating time 5s in binary control for SMA1. It obviously shows that the resistance
will increase first and then decrease during the heating process. When the power is
turned off, the resistance increases from t3=10s during the cooling process. Fig. 4-12
shows results of resistance with heating time 3s in binary control for SMA2 and the
resistance increases from t3=8s during the cooling process. It obviously shows the same
results as SMA1. Therefore, the hysteresis loop with heating time 5s and 3s are used as
major hysteresis loop to identify the phase resistance of SMA1 and SMA2, respectively.
Table 4- 2 Transformation resistances parameters
Resistances
of SMA1
Ras 1.535Ω
Resistances
of SMA2
Ras 1.445Ω
Raf 1.480Ω Raf 1.390Ω
Rms 1.470Ω Rms 1.360Ω
Rmf 1.515Ω Rmf 1.413Ω
Displacement
of SMA1
Max 0.84mm Displacement
of SMA2
Max 0.71mm
Min -0.08mm Min -0.02mm
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
71
Fig. 4-11 Results of resistance with heating time 5s for SMA1
in ambient temperature 24°C
Fig. 4-12 Results of resistance with heating time 3s for SMA2
in ambient temperature 24°C
0 20 40 601.3
1.4
1.5
1.6
1.65
0 20 40 601.3
1.4
1.5
1.6
1.65
Time (s)
Res
ista
nce
(Ω
)
Time (s)
Res
ista
nce
(Ω
)
Cooling
Heating
Cooling
Heating
t3
t3
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
72
Fig. 4-13 Major hysteresis loop for SMA1 in ambient temperature 24°C
Fig. 4-14 Major hysteresis loop for SMA2 in ambient temperature 24°C
1.4 1.45 1.5-0.2
0
0.2
0.4
0.6
0.8
1.3 1.35 1.4 1.45 1.5-0.2
0
0.2
0.4
0.6
0.8
Resistance (Ω)
A
Cooling
D C
B
Heating
Dis
pla
cem
ent
(mm
)
Rms Raf Rmf Ras
Resistance (Ω)
A
Cooling
D
C
B
Heating
Dis
pla
cem
ent
(mm
)
Rms Raf Rmf Ras
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
73
Fig. 4-13 and Fig. 4-14 show the phase resistance of major hysteresis loop of
SMA1and SMA2, respectively. In segment AB, the SMA wires extend quickly and
complete the phase resistances transformation from Rmf to Ras; in segment BC, the SMA
wires shrink and the resistances decrease from Ras to Raf; in segment CD, the resistances
decrease from Raf to Rms and the SMA wires cool but the output displacements remain
practically unchanged; According to the criteria mentioned in chapter 3, the resistance
decreases from Raf to Rms with 10% declines in the output strain. In segment DA when
the output displacements are less than 0.76mm and 0.64mm, the SMA wires start the
transformation from austenite to martensite for SMA1 and SMA2 at D, respectively.
More detailed information about the transformation resistances of SMA1 and SMA2 is
shown in Table 4-2.
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
74
4.4.2 Tuning the PID Parameters
The PRFC is proposed here to improve on the slow response speed obtained with
binary control. Fig. 4-6 shows the control arrangement of the PRFC where the
resistance coordinator sends the phase resistance as input to SMA1 and SMA2. The PID
controller should be tuned first to obtain proper parameters in the feedback control.
Fig. 4-15(i) shows the results of PID control using resistance as feedback to maintain
the resistance of SMA1 at Rms. When Kp=500, Ki=0, and Kd=0, the resistance increases
quickly since the input voltage which is used to heat and maintain the SMA wire is not
enough. As shown in Fig. 4-15(ii), this results in the output displacement decreasing
quickly without any maintaining; when Kp=4000, Ki=1, and Kd=0, the output
displacement is unstable because Ki=1 is too large and SMA is overheated, the
resistance cannot be maintained at Rms; when Kp=4000, Ki=0.5, and Kd=0, the resistance
can be maintained at Rms from 38s to 50s and there is no steady-state error e. However,
the output displacement (Fig. 4-15(ii)) is maintained at 0.83mm and there is long
latency duration Tm=12s; when Kp=4000, Ki=0.2, and Kd=0, the resistance is maintained
at Rms and the output displacement is maintained around 0.76mm from 23s to 50s even
if there is small steady-state error e. When the power is turned off from 50s, the output
displacement declines quickly. Therefore, Kp=4000, Ki=0.2, and Kd=0, are selected as
the parameters in the PID control. In addition, the PI control can meet the need in the
experiment. Therefore, parameter Kd is set to be zero in the PID controller.
According to the same method, the parameters for SMA2 are tuned. Fig. 4-16(i) and
Fig. 4-16(ii) show the results of PID control using resistance as feedback and output
displacement, respectively. It clearly shows that Kp=4000, Ki=0.18, and Kd=0, are the
suitable parameters to maintain the output displacement at 0.64mm.
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
75
Fig. 4-15 Results with different PID parameters for SMA1
in ambient temperature 24°C
0 20 40 60 80-0.2
0
0.2
0.4
0.6
0.8
1
1.2
0 20 40 60 801.2
1.3
1.4
1.5
1.6
1.7
kp=4000, ki=0.1, kd=0kp=500, ki=0, kd=0kp=4000, ki=0.2, kd=0kp=4000, ki=1, kd=0kp=4000, ki=0.5, kd=0Reference
40 45 50 551.441.461.481.5
1.52
e
Time (s)
Res
ista
nce
(Ω
)
Time (s)
Dis
pla
cem
ent
(mm
)
Maintaining
Maintaining
(i)
(ii)
Tn
Tm
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
76
Fig. 4-16 Results with different PID parameters for SMA2
in ambient temperature 24°C
0 20 40 60 801.2
1.25
1.3
1.35
1.4
1.45
1.5
1.55
1.6
Referencekp=4000, ki=0.18, kd=0kp=4000, ki=1, kd=0kp=500, ki=0, kd=0
0 20 40 60 80 100
0
2
4
6
0 20 40 60 80 1001.2
1.25
1.3
1.35
1.4
1.45
1.5
1.55
1.6
0 20 40 60 80
-0.2
0
0.2
0.4
0.6
0.8
kp=4000, ki=1, kd=0kp=4000, ki=0.18, kd=0kp=500, ki=0, kd=0
Res
ista
nce
(Ω
) D
ispla
cem
ent
(mm
)
Time (s)
Maintaining
Maintaining
Time (s)
(i)
(ii)
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
77
4.4.3 Results with PRFC
The two SMA wires will be actuated separately, using these parameters to test the
possibility of obtaining the results discussed in Fig. 4-4. Fig. 4-17 shows the results with
PRFC for SMA1 when the power for SMA2 is turned off. Fig. 4-17(i) shows, in the
heating process, the input voltage is Vh which can make a maximum output
displacement without overheating. In the maintaining process, the resistance is
maintained at Rms using resistance as feedback. Fig. 4-17(iii) shows the results of output
displacement. It clearly shows the heating, maintaining, and cooling processes and the
output displacement is maintained successfully during the maintaining section. Fig.
4-18 shows the results with PRFC for SMA2 when the power for SMA2 is turned off.
Fig. 4-18(i) shows that the input voltage is 5V and the heating time is 3s to obtain a
maximum output displacement during the heating process. Fig. 4-18(iii) shows the
results of output displacement of SMA2. During the maintaining section, the output
displacement is maintained at 0.64mm from 123s to 185s successfully.
With the results of SMA wires actuated separately, the two SMA wires will be
actuated together to demonstrate the results of rapid response speed with PRFC. Fig.
4-17(ii) and Fig. 4-18(ii) are plots of the variations of the actual resistances of SMA1
and SMA 2 using the phase resistances as the feedback. And the resistance is maintained
at Rms successfully. When the resistance of SMA2 reaches Raf at C=123s, SMA1 cools
quickly from the phase resistance Rms; then when the resistance of SMA1 reaches Raf at
E=185s, SMA2 cools quickly from the phase resistance Rms; during CE, there is no
output displacement though the resistances of SMA2 increase from Raf to Rms. The
latency duration of the cooling process is shortened, combining the heating process AC
of SMA2 and the cooling process CD of SMA1.
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
78
Fig. 4-17 Results with PRFC for SMA1 in ambient temperature 24°C
100 150 200 250 300
0
2
4
6
100 150 200 250 3001.3
1.4
1.5
1.6
1.68
100 150 200 250 300
-0.5
0
0.5
1
1.5
2
ResistanceReference
Res
ista
nce
(Ω
)
Time (s)
Volt
age
(V)
Dis
pla
cem
ent
(mm
) (i)
(ii)
(iii)
Vh
Cooing Maintaining
Cooing Maintaining Heating
Vm
C D
C D
E
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
79
Fig. 4-18 Results with PRFC for SMA2 in ambient temperature 24°C
100 150 200 250 300
0
2
4
6
100 150 200 250 3001.3
1.4
1.5
1.6
100 150 200 250 300
0
1
2
ResistanceReference
Res
ista
nce
(Ω
)
Time (s)
Volt
age
(V)
Dis
pla
cem
ent
(mm
)
(i)
(ii)
(iii)
Vh
Maintaining Cooling
Maintaining Cooling Heating
Vm
C E A
C A
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
80
The output displacement is illustrated in Fig. 4-19. It shows that the maximum output
displacement (1.47mm) is the sum of both SMA wires. Part 1 and part 2 are the
available results of cycle for SMA2 and SMA1, respectively. The detailed output
displacement is shown in Fig. 4-20 with two cycles.
Fig. 4-19 Total output displacement with PRFC in ambient temperature 24°C
Fig. 4-20 Detailed output displacement with two cycles for PRFC
in ambient temperature 24°C
100 150 200 250 300
-0.5
0
0.5
1
1.5
2
120 140 160 180 200 220
0.6
0.8
1
1.2
1.4
1.6
Time (s)
Dis
pla
cem
ent
(mm
)
Part 1 Part 2
Time (s)
Dis
pla
cem
ent
(mm
)
Part 1 Part 2
Total
Available
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
81
4.5 Consideration and Discussion
The arrangement proposed here improves the response speed of the thick SMA wires
significantly in comparison with the binary control. There are , however, a few critical
points needs further consideration.
1. Since the maximum output displacement of both SMA wires is different. To
obtain clear and accurate results, part 1 will be selected to compare with binary
control of SMA2, and part 2 will be compared with binary control of SMA1. As
illustrated in Fig. 4-21, the output displacement with PRFC starts to extend at 10s
during the cooling process. According to the criteria in chapter 3, when the output
displacement is less than 90% of the maximum displacement, the SMA1 starts the
transformation from martensite start phase to martensite finish phase at 13s, and
the latency duration with PRFC TLp=3s. The SMA1 completes the phase
transformation at e=46.5s. Fig. 4-22 shows the results with PRFC for SMA2. The
SMA2 completes the phase transformation at e=43s. More detailed information is
listed in Table 4-3. The average cycle time for the binary control and proposed
method is 54.5s and 39.8s, respectively. The response improvement is given by
%1.275.54
7.14
5.54
8.395.54
(4-8)
Table 4-3 Parameters of the SMA1 and SMA2
Binary control PRFC
Tcd Cycle time TLp Cycle time
SMA1 13s 54s 3s 41.5s
SMA2 14s 55s 2s 38s
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
82
Fig. 4-21 Results with the PRFC and binary control for SMA1
in ambient temperature 24°C
Fig. 4-22 Results with the PRFC and binary control for SMA2
in ambient temperature 24°C
0 10 20 30 40 50 60-0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Binary controlProposed
0 10 20 30 40 50 60-0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Binary controlProposed
Time (s)
Tcd
TLp
Dis
pla
cem
ent
(mm
)
Time (s)
Dis
pla
cem
ent
(mm
)
Offset
Offset
Cooling
Heating
Cooling Heating
SMA1 SMA2
e f
f
e
Tcd
TLp
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
83
2. Theoretically, the MOSFET should be turned off during the cooling process in
order to obtain rapid cooling response speed. One limitation is that the
microcontroller needs time to read the analog inputs voltage V1, V2 and V3 to
calculate the resistance by turning on the MOSFET and turning off it when the
reading is finished, which makes longer time for the SMA finishes phase
transformation from austenite to martensite. Fig. 4-23 shows the results with or
without reading the input voltage to calculate the resistance with ambient
temperature 24°C. There will be small current through the SMA wire by turning
on MOSFET when reading the analog input voltage, resulting in SMA finishes
the martensite phase at f2=71s instead of f1=68s. It means that the cooling speed
will be slower by calculation the resistance of SMA during cooling process even
if the heating speed is the same. Therefore, when the output displacement with
binary control is detected by displacement sensor, the resistance is calculated as
well by reading the analog inputs in order to compare the output displacement
with proposed method. In addition, as shown in Table 4-2, there is small
difference for the resistance of SMA wires because the calculation of resistance is
affected by the microcontroller, specimen, driver circuit and power source.
3. As shown in Fig. 4-21 and Fig. 4-22, the offset of two cycles during cooling
process is 0.05mm and -0.08mm for SMA1 and SMA2 respectively. The output
displacement (Fig. 4-21) is from SMA1 and SMA2 during the heating and
cooling processes for the proposed method, respectively. Since the maximum
output displacement of SMA2 is 0.71mm, the output displacement with PRFC for
SMA1 declines from 0.81mm and can not reach to zero position during the
cooling process. Then, the offset is positive; since the maximum output
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
84
displacement of SMA1 is 0.81 mm, the output displacement with PRFC for
SMA2 (Fig. 4-22) declines from 0.71mm and the offset is negative. In addition,
the insulation joint which is used to prevent short circuiting needs to be tight
enough to connect SMA wires to prevent offset caused by loosening.
4. As shown in Fig. 4-24, the results of output displacement are very similar with
heating time 5s for 5 tests, however, the offset for maximum output displacement
and martensite finish displacement are 0.02mm and 0.03mm, respectively. It
means there will be small offset for output displacement in each test with the
same heating and cooling time because the output displacement is affected by
hysteresis effect and sensitive to ambient environment. Therefore, the offset for
the output displacement is unavoidable.
Fig. 4-23 Binary control with or without resistance calculation
in ambient temperature 24°C
0 20 40 60-0.5
0
0.5
1
1.5
Time (s)
Dis
pla
cem
ent
(mm
)
With reading
Without reading
f1 f2
Heating
Cooling
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
85
Fig. 4-24 Results with the binary control for 5 tests
in ambient temperature 24°C
5. For the two cycles shown in Fig. 4-20, the maximum output displacement is
based on one single SMA wire even though two SMA wires are used in the
experiment with PRFC. The total output displacement is 1.47mm, however the
effective output displacement is 0.71mm and 0.81mm. Therefore, the results of
rapid response speed is achieved on the cost of maximum output displacement
10 150.75
0.8
0.85
0.9
0.95
60 65
-0.1
-0.05
0
0.05
0 20 40 60
0
0.5
1
1.5
Time (s)
Dis
pla
cem
ent
(mm
)
Offset Offset
Heating
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
86
which is a limitation for the proposed method.
6. The results only show the improvement of cooling speed with shortening the
latency duration Tcd , the latency duration Tab still exists even though is small
compared with Tcd. As shown in Fig. 4-15, there is still small hysteresis duration
Tn when the output displacement is maintained at martensite start displacement
during the cooling process when the power is turned off. Moreover, there is short
latency duration TLp with the PRFC. It means the latency duration caused by
hysteresis effect cannot be eliminated totally. But it does not deny the
effectiveness of the proposed method.
7. Comparing with the research in reference [5], there is no limitation caused by
Peltier device. The advantages are as follows.
(i) There is no need to heat the Peltier in order to heat SMA, resulting in
saving energy.
(ii) Since the two SMA wires are connected by insulation joint, there is no
heat transfer between adjacent segments.
(iii) The SMA wires will not shift to adjacent units, as the SMA wire shrinks
and expands. Therefore, there is no error caused by adjacent segment.
(iv) At last, the experimental apparatus is light and convenient because of
without Peltier models in the apparatus.
4.6 Chapter Summary
An SMA actuator structure using two connected SMA wires which is able to generate
rapid response for thick SMA actuators is proposed. The results with the proposed
PRFC (Phase resistance feedback control) using the concept of phase transformation
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
87
resistances that divided the hysteresis loop into four parts, enable a shortening of
latency duration with two connected SMA wires coordinated to use the phase
resistances as feedback than with the binary control. Experimental results also
demonstrated that the average cycle time was 27.1 percent shorter than with binary
control.
To accurately identify phase resistances, experiments have shown that it is important
to determine the major hysteresis loop. It is also important to avoid overheating the
SMA wires by controlling the heating time which could otherwise slow the response
speed greatly.
CHAPTER 4: PHASE RESISTANCE FEEDBACK CONTROL TO ACHIEVE RAPID
RESPONSE SPEED
88
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
89
CHAPTER 5:
IMPROVEMENT OF RESPONSE SPEED
USING PHASE RESISTANCE AND
DISPLACEMENT AS FEEDBACK
5.1 Introduction
There are many researches on improvement of the response speed in the past as
mentioned in chapter 3 and chapter 4 [74, 80]. Conventional SMA actuator systems
consist of heating the entire length of SMA wire with electric current and cooling with
natural convection [62, 67], the wire shrinks or extends at the same time. There are long
latency durations for thick SMA wires, the response speed is slow which has been
discussed in chapter 4. In contrast, an approach, which separately controls two
connected SMA wires, making individual SMA wire can shrink or extend at different
time with the same maximum output displacement as an entire SMA wire, is proposed
in this chapter. The motivations will first be explained. In section 5.3 and 5.4, the
method and results will be described. Some discussions and conclusions about the
results are also presented in section 5.5 and 5.6, respectively.
5.2 Motivation and Target
In chapter 4, the PRFC demonstrates more rapid response speed than binary control.
However, the maximum output displacement is only detected from one single SMA wire.
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
90
Here, an alternative control method, phase resistance and displacement feedback control
(PRDFC), combining both the phase resistance and displacement as feedback, is
proposed to obtain rapid response speed as well. It minimizes cooling time by
shortening the long latency duration of thick SMA wires and the total output
displacement is the same as traditional control method. Experimental results here show
that rapid response speed is achieved using this method in comparison with the case in
which only displacement is used as feedback.
5.3 Method
5.3.1 Phase Resistance with Displacement Feedback Control (PRDFC)
Traditionally, a data collected from a displacement sensor can be used as feedback for
only one SMA wire in a position control. It is difficult to control the output
displacements of two SMA wires at the same time when they shrink or extend randomly
because they will interfere with each other. To prevent the two SMA wires from mutual
interference and obtain accurate position control results, a few technical issues must be
considered:
1. When the SMA wire completes the transformation from martensite to austenite, it
needs to be maintained in martensite starting phase which is ready to extend
without any changes in the output displacement;
2. When one SMA wire completes the transformation from martensite to austenite, the
other needs to start the phase transformation quickly to guarantee continuity of
output displacement, or vice-versa.
The critical aspect of the PRDFC method is to shorten the total cooling time by
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
91
switching the reference input from phase resistances to displacement with two
connected SMA wires that are coordinated, or vice-versa.
To justify the concept of PRDFC, a step signal is used as reference input for a
position control of SMA actuator first. Fig. 5-1 shows the basic theory behind this
method, using the definitions in Fig. 4-2, Fig. 4-3 and Table 4-1, and it can be explained
as follows.
Fig. 5-1 Schematic of step reference for the PRDFC method
As shown in Fig. 5-1(i), the step signal is divided into two parts, part 1 is used as
reference input signal for SMA1, part 2 is used as reference input signal for SMA2.
Since the data collected from the displacement sensor can be used as feedback for only
one SMA wire, the SMA wire need to be maintained at Rms to prevent from any changes
in the output displacement when the other one extends or contracts. As shown in Fig.
5-1(ii), the reference input switches from displacement to resistance for SMA1 at b, and
a
Dis
pla
cem
ent
Time b c d
Res
ista
nce
(i)
(ii)
Part2
Part1 SMA1
SMA2
SMA1
D1
D2
0
Rms
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
92
during section bc the output displacement is maintained. This means that only one SMA
wire provides the output displacement in each part.
During section ab, part 1 is used as the reference displacement for SMA1, and once
the total output displacement reaches D1 at b, the reference input quickly switches from
displacement to resistance and the resistance of SMA1 is maintained at Rms. The result
is that there is no displacement output from SMA1 during section bc.
Fig. 5-2 Schematic of ramp reference for the PRDFC method
During section bc, part 2 is used as reference displacement input for SMA2 from b,
immediately. The data collected from the displacement sensor can be used as feedback
to control the position of SMA2 because SMA1 is maintained in martensite starting
phase and it is ready to extend without any changes in the output displacement.
During section cd, once the total output displacement reaches D2 at c, both the power
for SMA1 and SMA2 are turned off to achieve quickly cooling speed. Since SMA1 has
a
Dis
pla
cem
ent
Time b c d
Res
ista
nce
(i)
(ii)
Part2
Part1
SMA1
SMA2
SMA1
D1
D2
0
Rms
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
93
completed the phase transformation from austenite to matensite starting during section
bc, the latency time Tcd of SMA1 is shortened, making it possible to obtain fast cooling
response speed during section cd. In order to obtain the maximum output displacement
of SMA continuously, the reference input needs to increase step by step. Since the
sampling time is 0.05s, the time of each step for reference input needs to be less than
sampling time. In this chapter, the ramp signal is used in the position control to get the
maximum output displacement. Fig. 5-2 shows the basic theory behind this method as
well, using ramp signal as reference input. Part 1 is used to obtain the maximum output
displacement of SMA1. Then, the Rms of SMA1 can be used as feedback to maintain the
output displacement of SMA1.
5.3.2 Control System
Fig. 5-3 Block diagram of PRDFC, (i) Displacement feedback control; (ii) Phase
resistance feedback control
As discussed in the introduction section, many efforts have been made to model and
control behavior of SMA actuators. In this section, a new control system combined by
Displacement Displacement
PWM SMA
Resistance
Rms
e +
-
-
(i)
(ii)
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
94
displacement and resistance feedback is proposed here to apply the PRDFC method to
control SMA actuator. Fig. 5-3(i) and Fig. 5-3 (ii) show the displacement and resistance
feedback system, respectively. The control system switches from resistance to
displacement feedback control for the position control of SMA1 or SMA2; the control
system switches from displacement to resistance feedback control to maintain the output
displacement of SMA1 or SMA2. The reason that switches control system between
displacement feedback control and resistance feedback control is that there is no
displacement output from SMA wires by maintaining the resistance of SMA wires at Rms
when they complete phase transformation from martensite to austenite, making it
possible to obtain fast response speeds during the cooling processes by shortening the
latency duration without reducing the accuracy of position control.
Fig. 5-4 Schematic of the experimental setup for displacement feedback control
For the displacement control of SMA1 or SMA2, the displacement is measured by
the displacement sensor and resistance is calculated by Eq. (3-1), using the data sent by
microcontroller. To be able to use the phase resistance and displacement as feedback, a
PI controller is installed. The PI controller is denoted by PWM which is sent to
MOSFET and converted to heat voltage. It can be calculated by Eq. (4-6) and Eq. (4-7).
L=LSMA1+LSMA2
Bias spring
Laser
sensor
sensor
Reflector
Load cell
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
95
The experimental setup is the same as Fig. 4-7 with PRDFC. To be able to compare
PRDFC with traditional displacement feedback control, it is necessary to use the same
length of SMA wire by sending the same PWM value (calculated by PID controller in
Fig. 4-8) to heat the SMA wires. Then, as shown in Fig. 5-4, the length of SMA wire
with traditional displacement feedback control, L, is expressed by
L=LSMA1+LSMA2 (5-1)
where LSMA1 and LSMA2 are the length of SMA1 and SMA2, respectively.
5.4 Results
5.4.1 Tuning PID Parameters
In order to obtain good results with PRDFC, the parameters of resistance feedback in
the PID controller based on Eq. (4-6) are the same as used in chapter 4. For the
parameters of displacement feedback, it needs to be tuned again. Fig. 5-5 shows the
results of PI control using displacement as feedback. When Kp=500 and Ki=0, the
maximum output displacement is small and there are large steady-state error since the
input voltage which is used to heat the SMA wire is small. When Kp=1000 and Ki=0, the
maximum output displacement increases. However, there is still steady-state error.
When Kp=1000 and Ki=0.2, the tracking performance is the best. With Kp=1000 and
Ki=1, the output displacement is the worst since the input voltage changes largely
shown in Fig. 5-6.
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
96
Fig. 5-5 Results with different parameters of PI controller
in ambient temperature 24°C
Fig. 5-6 Input voltage for different parameters of PI controller
in ambient temperature 24°C
0 20 40 60 80-0.5
0
0.5
1
1.5
kp=500, ki=0kp=1000, ki=0kp=1000, ki=0.2kp=1000, ki=1
0 20 40 60 800
1
2
3
4
5
6
kp=500, ki=0kp=1000, ki=0kp=1000, ki=0.2kp=1000, ki=1
Time (s)
Dis
pla
cem
ent
(mm
)
Time (s)
Volt
age
(V)
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
97
According to the method shown in Fig. 5-1, experiments are conducted with step
signal as reference input. The maximum output displacement of step is 1.4mm, and then
the maximum output displacements for SMA1 wire are 0.8mm (0.01mm offset from the
maximum output displacement 0.81mm) which leads to a major hysteresis loop of
displacement-to-resistance, making it possible to use the phase resistances as feedback
during the phase resistance feedback control. As shown in Fig. 5-7, the resistance of
SMA1 decreases from 10s since the displacement is used as feedback from 10s to 50s.
From 50s to 90s, the resistance of SMA1 is maintained at Rms . As shown in Fig. 5-8, the
output displacement is unchangeable and maintained at 0.74 mm during the maintaining
period and the SMA1 completes the cooling process at f=135s. Therefore, it is possible
to actuate SMA2 using the displacement as feedback from 50s to 90s because the
SMA2 will not be interfered by the output displacement of SMA1.
Fig. 5-7 Results of resistance for SMA1 with PRDFC in ambient temperature 24°C;
resistance control: Kp=4000, Ki=0.2
0 50 1001.3
1.4
1.5
1.6
1.7
1.8
1.9
2
ResistanceReference
Time (s)
Res
ista
nce
(Ω
)
Maintaining
Cooling
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
98
Fig. 5-8 Results of output displacement for SMA1 with PRDFC
in ambient temperature 24°C
5.4.2 Results with PRDFC
The experimental results of the proposed PRDFC and conventional displacement
feedback control (based on Eq. (4-6) with Kp=1000 and Ki=0.2 as parameters, e(t) is the
signal error of displacement) are shown in Fig. 5-9. And the traditional method leads to
latency more latency duration (Tcd=5.5s) than proposed method, which is caused by
hysteresis effect of SMA (same phenomenon observed in Fig. 3-9). It is observed that
the proposed PRDFC behaves more rapid response speed than the traditional
displacement feedback control during the cooling process from 90s to 150s, though both
systems in this test have similar tracking trajectory during the heating process from 10s
to 90s. Since during tracking period of SMA2 from 50s to 90s, the latency duration of
SMA1 is shortened, which leads to rapid cooling speed during the cooling process from
90s when the power is turned off. According to criteria to decide the martensite finish
time with horizontal line mentioned in section 3.5, the proposed and traditional method
complete the cooling process at e=138s and f=148s, respectively.
0 50 100 150 200-0.5
0
0.5
1
ResistanceReference
Time (s)
Dis
pla
cem
ent
(mm
) Maintaining
Cooling
f
0.74mm
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
99
Fig. 5-9 Results of PRDFC and traditional method in ambient temperature 24°C;
displacement control: Kp=1000 and Ki=0.2; resistance control: Kp=4000 and Ki=0.2
Fig. 5-10 Results of resistance for SMA2 with PRDFC in ambient temperature24°C;
displacement control: Kp=1000 and Ki=0.2
0 50 100 150-0.5
0
0.5
1
1.5
ProposedTraditionalReference
0 50 1001.3
1.4
1.5
1.6
1.7
1.8
1.9
2
Time (s)
Dis
pla
cem
ent
(mm
)
e f
Time (s)
Res
ista
nce
(Ω
)
Heating Cooling
Heating
Part 1
Part 2
0.06mm
Tcd
TLp
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
100
Fig. 5-11 Results of PRDFC and traditional method in ambient temperature 24°C;
displacement control: Kp=1000 and Ki=0.2; resistance control: Kp=4000 and Ki=0.2
Fig. 5-12 Results of resistance for SMA1 with PRDFC in ambient temperature 24°C;
resistance control: Kp=4000, Ki=0.2
0 50 100 150-0.5
0
0.5
1
1.5
ProposedTraditionalReference
0 50 1001.35
1.45
1.55
1.65
1.75
1.85
1.952
ResistanceReference
Time (s)
Res
ista
nce
(Ω
)
Maintaining Cooling
Heating
Part 1
Part 2
o
Time (s)
Dis
pla
cem
ent
(mm
)
e f
Heating
-0.07mm
Tcd
TLp
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
101
Fig. 5-13 Results of resistance for SMA2 with PRDFC in ambient temperature 24°C;
displacement control: Kp=1000 and Ki=0.2
Fig. 5-10 shows the results of resistance for SMA2. It also demonstrates the variation
of resistance using displacement as feedback during heating process. When the power is
turned off from 90s, the resistance increases quickly during the cooling process.
In addition, experiments are also conducted with ramp signal using the same PI
controller as step as reference input to demonstrate the feasibility of the proposed
method. Fig.5-11 shows the experimental results of the proposed PRDFC and traditional
control with ramp signal. Part 1 and part 2 are used as displacement feedback for SMA1
and SMA2, respectively. The tracking trajectory of PRDFC is also the same as that of
displacement feedback control plotted using the black line. It is observed that the
proposed PRDFC behaves more rapid response speed than the traditional method during
the cooling process from 45s to 130s as well. Since during tracking period of SMA2
from 30s to 45s, the latency duration of SMA1 is shortened, which leads to rapid
0 50 100 1501.3
1.4
1.5
1.6
1.7
1.8
1.9
2
Time (s)
Res
ista
nce
(Ω
)
Cooling Heating
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
102
cooling speed during the cooling process from 45s when the power is turned off. And
the traditional method leads to more latency duration (Tcd=6s) than proposed method.
According to criteria to decide the martensite finish time with horizontal line mentioned
in 3.5 as well, the proposed and traditional method complete the cooling process at
e=108s and f=120s, respectively.
Fig. 5-12 and Fig. 5-13 show the results of resistance with PRDFC. In Fig.5-12, the
resistance of SMA1 decreases from 10 s since the displacement is used as feedback
from 10s to 30s. During this heating process, the SMA1 completes the transformation
from martensite to austenite. From 30s to 45s, the resistance of SMA1 is maintained at
Rms, which makes the output displacement unchangeable during the maintaining period
to actuate SMA2 using the displacement as feedback because the SMA2 will not be
interfered by the output displacement of SMA1. During maintaining process, the SMA1
completes the transformation from austenite finish phase to martensite start phase which
leads to a shortening latency duration Tcd.
5.5 Consideration and Discussion
The proposed method demonstrates more rapid response speed than traditional
displacement feedback control. There are, however, a few critical points needing further
consideration.
1. As shown in Fig. 5-9 and Fig. 5-11, both the step and ramp signals show that the
proposed method lead to more rapid response speed than traditional displacement
feedback control because the proposed method shortens the latency duration Tcd
of SMA1 which is caused by the hysteresis effect. However, as mentioned in
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
103
section 4.5, there is still latency duration TLp which can not be eliminated totally.
In addition, the parameters of tuning PI controller is important since the best and
similar position tracking can eliminate the possibility that the rapid speed of
proposed method arises from other factors. Another limitation is only the latency
duration of SMA1 is shortened in these tests even though two SMA wires are
used in the experiments.
2. As shown in Fig. 5-9 and Fig. 5-11, the speed of heating process with traditional
method for step input 0.8mm is faster than with the proposed method. Because
the SMA length for the traditional method is 280 mm, however, the SMA length
for the proposed method is 140mm for part 1. With the same PID parameter and
reference input, the maximum output displacement for the traditional method is
larger than proposed method in part 1, resulting faster response speed during
heating process. Since the remaining maximum output displacement of both
methods is the same for the part 2, say 0.71mm, the speed of heating process is
almost the same.
3. Concerning about PRDFC, it is important to make two SMA wires output
displacement continuously. Once there is no output displacement from SMA1,
SMA2 should be actuated quickly. The part o caused by thermal expansion of
SMA2 (Fig. 5-11) leads to less accurate for the position tracking control
compared with tradition method. In addition, it is important to distinguish the
maximum output displacement of SMA wire to avoid overheating with PRDFC.
A maximum ramp signal, 1.6mm(1.0mm for SMA1, 0.6mm for SMA2), is used
as reference input to verify this statement. As shown in Fig. 5-14, part 1 and part
2 are used as displacement feedback for SMA1 and SMA2, respectively. When
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
104
the maximum reference input for the SMA1 is 1.0mm, but the maximum output
for SMA1 is only m=0.8mm (0.01mm offset from the maximum output
displacement 0.81mm), then SMA1 will be over heated from m to n, which leads
to more overshoot and undershoot than traditional method from 30s to 40s.
4. As shown in Fig. 5-9 and Fig. 5-11, the maximum offset for the output
displacement is respective 0.06mm and -0.07mm when SMA finishes the
transformation from austenite to martensite during the cooling process at f and e,
which is arisen from the same reason mentioned in section 4.5.
Fig. 5-14 Results of overheating for SMA1 in ambient temperature 24°C;
displacement control: Kp=1000 and Ki=0.2; resistance control: Kp=4000 and Ki=0.2
0 50 100 150-0.5
0
0.5
1
1.5
2
Proposed TraditionalReference
30 40 50 60
0.8
1
1.2
n m
Time (s)
Dis
pla
cem
ent
(mm
)
Part 1
Part 2
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
105
5.6 Chapter Summary
An SMA actuator structure and control system using two connected SMA wires
which is able to generate rapid response for thick SMA actuators is proposed. Two sets
of signal are used as reference input to test the results of both proposed and traditional
method. The results show that using only displacement as feedback leads to slower
cooling speeds than with the proposed PRDFC.
Since experiments demonstrated that the thick SMA wires suffer from significant
hysteresis effects with long latency duration (chapter 3), to obtain a rapid response of
the SMA actuators, this chapter proposes a PRDFC (Phase resistance with displacement
feedback control) method using the concept of phase transformation resistances that
dividing the hysteresis loop into four parts (chapter 4).
In the experiments, SMA1 is used in the position tracking control with displacement
as feedback first. Then SMA2 is used in the position tracking control when SMA1 is
maintained using resistance as feedback. This enables a shortening of the latency
duration with two connected SMA wires coordinated by using the phase resistance and
displacement as feedback when compared with the case where only displacement
feedback is used.
CHAPTER 5: IMPROVEMENT OF RESPONSE SPEED USING PHASE
RESISTANCE AND DISPLACEMENT AS FEEDBACK
106
.
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
107
CHAPTER 6:
MODELING OF AN SMA ACTUATOR BASED
ON THE LIANG AND ROGERS MODEL
6.1 Introduction
In modeling the hysteresis of an SMA to accurately estimate the behavior of the SMA
actuator, many models have been proposed during the last decade. However, the main
difficulty in modeling SMA actuators is due to the non-linear saturated hysteresis effect
during the phase transformation from martensite to austenite, or vice-versa. Hirose,
Ikuta and Umetani proposed a two-phase model for SMA using the sub-layer model, a
commonly used method to describe nonlinear stress-strain relationships in solid
mechanics [86]. Tanaka developed a thermo-mechanical law that governs the
stress-strain behavior of SMA elements [30]. Williams and Mohammad introduced a
model of an SMA actuator which consists of four sub-models: a heat transfer model, an
SMA thermo-mechanical model, a phase transformation kinetics model, and a
dynamic/kinematic model [87]. However, these models did not show the minor
hysteresis relationship between input and output even though they can precisely
reproduce the thermo-mechanical behavior of SMA. Dutta and Ghorbel proposed a
model using a differential hysteresis model capable of representing the major and minor
hysteresis loops [38]. However, this model can not show the relationship between
martensite start temperature and the maximum temperature during heating process of
each minor hysteresis loop. In this chapter, the motivations will first be explained. In
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
108
section 6.3 and 6.4, the method and results will be described. Some discussions and
conclusions will be shown in section 6.5 and 6.6.
6.2 Motivation and Target
In this chapter, a successful empirical relation proposed by Liang and Rogers is
introduced in order to model the major hysteresis behaviors of SMA actuators, which
represents the amount of austenite fraction transformed on a temperature [34]. Liang
and Rogers described a unified, thermomechanical constitutive model to quantitatively
predict the stress-strain relation and shape memory behavior. It also focused on shape
memory effects and provided a theoretical guide to the design of SMA based on
intelligent material and structures. However, it is difficult to apply the method to
practical application without representing the minor hysteresis loops, successfully.
Based on the empirical relation of the Liang and Rogers model, a modified Liang and
Rogers model is demonstrated to consider the major and minor hysteresis behaviors. An
experimental setup used for verification of the modeling system is presented and some
test series are conducted to identify the parameters of the modified Liang and Rogers
model. The reasonable agreement achieved between curves predicted by the modified
Liang and Rogers model and the measured data shows that the proposed model is
efficient in modeling the hysteresis of the SMA actuator system.
6.3 Method
6.3.1 Thermal Model of Heat Transfer and Temperature
As mentioned in chapter 2, an SMA returns to some predefined shape or size when
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
109
subjected to an appropriate thermal procedure, and is known as a shape memory effect.
The shape memory effect arises from temperature and stress dependent shifts in the
crystalline structure of the SMA, shifts between martensite and austenite phases. The
system gains heat energy from an electrical current, and loses part of it to the
environment. As suggested in Fig. 6-1, this model considers an SMA element as a
three-element system in which thermal energy is concerted into a phase transformation
and then into mechanical work. In this section, a new mathematical model of the SMA
actuator is introduced, including the major and minor hysteresis loops to model the
transformation model of SMA actuator.
Fig. 6-1 Block diagram model of SMA
In this section, the heat transfer problem of the SMA wire is described. The balance
of the heat energy governs the temperature of the SMA actuator. For a spring-biased
SMA wire, the thermal model between input voltage V and the output temperature T ,
which is shown in module 1 (Fig. 6-1), is a first-order system given by
)()(
400
20
2
0
ambTThLdR
tV
dt
dTLdc
(6-1)
Module 1 Module 2
Module 3
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
110
where is the density of the SMA; c is the specific heat coefficient; 0L is the length
of the SMA wire; 0d is cross-sectional diameter of the SMA; V is the voltage applied; R
is the resistance; h is the convection heat transfer coefficient, and ambT is the ambient
temperature.
Fig. 6-2 Schematic of the input voltage
Since the actual temperature of the SMA wire is not measured in the experiments,
direct validation of the heat transfer model mentioned above is not possible. However,
the simulation result is able to demonstrate the thermal transfer model. Fig. 6-2 shows
that six triangular voltages are set as the input voltage for simulation. The maximum of
input voltage is 2V for the major loop and the others are for the minor loops. Fig. 6-3
shows the simulation results of the output temperature obtained from Eq. (6-1)
corresponding to the input voltage shown in Fig. 6-2.
0 200 400 600 8000
0.5
1
1.5
2
Volt
age
(V)
Time (s)
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
111
6.3.2 Phase Transformation and Mechanical Model
During the heating process, a phase transformation occurs from martensite to
austenite, while during the cooling process, the opposite transformation occurs, and
SMA wires show a hysteresis effect during both phase transformations. The extent of
austenite to martensite tansformation is characterized by the martensite fraction m .
Martensite fraction is defined as the volume fraction of M phase present in the SMA at
any instant. Therefore, 10 m . As shown in Fig. 6-4, the transformation is
characterized by the initial and finish temperatures. MST and MFT are the initial and
final temperatures of martensite, while AST and AFT are the initial and final
temperatures of austenite, respectively.
Fig. 6-3 Schematic of the output temperature
0 200 400 600 80020
30
40
50
60
70
80
90
Time (s)
Tem
per
ature
(
)
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
112
Fig. 6-4 Schematic of martensite fraction-temperature hysteresis
The classical Liang and Rogers model uses trigonometric functions to characterize
the hysteresis effect of module 2 for the SMA actuators. During the heating and cooling
processes, the functions of the martensite fraction mh and mc are expressed by
2
21
1
0
)(cos15.0
1
MT
MTM
MT
TT
CA
FTT
ASAF
A
AS
mh
(Heating) (6-2)
Heating
Cooling
Temperature
Mar
tensi
te f
ract
ion
0
1
TMF TAS TMS TAF
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
113
4
43
3
0
)(cos15.0
1
MT
MTM
MT
TT
CA
FTT
MFMS
M
MF
mc
(Cooling) (6-3)
where M1= TAS +F/(A∙CA); M2= TAF +F/(A∙CA), M3= TMF+F/(A∙CM); M4= TMS +F/(A∙CM);
T is the temperature of the SMA wire; F is the pre-tension; A is the cross section area of
the wire; CM is the stress rate of martensite, and CA is the stress rate of austenite.
Then the austenite fractions during the heating and cooling processes can be defined
as mhah 1 and mcac 1 , respectively. Since the effect of pre-tension is small
in the experiments, the equations proposed by Liang and Rogers that model such
transformations as function of temperature are simplified as
AS
AFAS
AF
ASAF
ASah
TT
TTT
TT
TT
TT
0
)(cos15.0
1
(Heating) (6-4)
MF
MSMF
MS
MFMS
MFac
TT
TTT
TT
TT
TT
0
)(cos15.0
1
(Cooling) (6-5)
where ah and ac are the amount of austenite fractions during the heating and cooling
processes, respectively.
However, the Liang and Rogers model only represents the major hysteresis loop of
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
114
phase transformation without considering the martensite starting temperature of minor
loops. A typical austenite fraction-temperature hysteresis schematic is shown in Fig. 6-5
[38]. The hysteresis loop corresponding to complete phase transformation is called the
major hysteresis loop, while incomplete phase transformation yields minor hysteresis
loops within the major hysteresis loop. The martensite starting temperature of each
minor loop is the same value as the major loop [38].
According to the experimental data described in section 6.4, the minor hysteresis
loops can be revised in a new shape shown in Fig. 6-6. The martensite starting
temperature of each minor loop is different. The solid line represents the major
hysteresis loop and the dashed lines are for the minor loops. Since the transformation
temperature variation caused by the applied stress is small in this experiment, we
assume that the phase transformation temperatures are constant throughout for major
and minor hysteresis loops.
If ASi TT , then )(Tahi =0, and therefore, only the case of ASi TT is considered.
For a hysteresis loop, the austenite fraction ahi during the heating process is
expressed by
ASi
AFiAS
AFi
ASAF
ASahi
TTT
TTTT
TTT
TT
TTT
0
)(cos15.0
1
)( ),,2,1( Ni (6-6)
where Ni ,,3,2,1 ; iT is the maximum temperature during the heating process of a
hysteresis loop, including minor hysteresis loops. However, Eq. (6-6) is derived from
Eq. (6-4) without modified minor hysteresis loops of heating process.
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
115
Fig. 6-5 Schematic of typical austenite fraction-temperature hysteresis
Fig. 6-6 Schematic of austenite fraction-temperature hysteresis with modification
Temperature
Fra
ctio
n
0
1
TMF TMS TAS TAF
Heating
Cooling
Aust
enit
e fr
acti
on
Temperature
0
1
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
116
When minorfminors TTT i , then the austenite fraction increases first, and is
maintained unchanged, then decreases during the cooling process; the equation of the
increase and maintaining stages acii is expressed by
iiMSiiiahi
iiiiiahi
acii TTTTzT
TTTTzTT
ii)()(
)()()(
(6-7)
where iiT is the starting temperature of the maintaining stage and can be expressed by
iiiiiiiiii TTTNTT minminmax )( , iiiiii TTT maxmin (6-8)
where iiTmax and iiTmin are the maximum and minimum of iiT . iiNT can be expressed
by
iiNT0
1
1
2
2
3
3
4
4
5
5 aNTaNTaNTaNTaNTa iiiii (6-9)
where iNT can be expressed by
minorsminorf
minors
TT
TTNT i
i
, fi TTT minorminors (6-10)
where minorsT and minorfT is the starting and finishing temperature with increase stages
of cooling process, respectively. MSiT is the martensite starting temperature of a minor
hysteresis loop; )(Tz i is defined as the function for the cooling process in the minor
hysteresis loop.
In order to make sure 1)(0 Tacii and simplify the equation of )(Tz i , the
temperature is normalized. )(Tz i can be expressed by
increase
maintaining maintaining
maintaining
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
117
iiMSiii
iii
iii
ii
i
TTTjTz
TTTTT
TTj
Tz
ii)(
)(
)( (6-11)
where ij is the constant for a minor hysteresis loop.
The martensite starting temperature MSiT , which must be determined to model minor
hysteresis loop, is expressed by
minorsminorsminorf )( TTTNTT MSiMSi , minorfminors TTT i (6-12)
where the normalized resistance MSiNT is expressed by
MSiNT0
1
1
2
2
3
3
4
4 mNTmNTmNTmNTm iiii (6-13)
where 4m , 3m ,
2m ,1m and 0m are the parameters obtained by polyfit of Matlab.
When minors1 TTT i and Ni TTT minorf , then 0)( Tz i , there are only
maintaining stages. Therefore, the martensite starting temperature MSiT can be
expressed by
eTqTiMSi ahi )( (6-14)
where q and e are the constants.
During the cooling process, the austenite fraction for a loop can be expressed by
MF
MSiMF
MSii
MSiacii
MFMSi
MF
acii
aci
TT
TTT
TTT
TTT
TT
T
T
)(
0
)(cos15.0
)(
)(
(6-15)
For the mechanical model, the module 3 (Fig. 6-1) shows the output displacement D,
which combines the thermal model and phase transformation model developed above, is
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
118
expressed by
coolingg
heatinggD
aci
ahi
(6-16)
where g is constant parameter.
With the three models mentioned above, the hysteresis loops of input voltage and
output displacement including the major and minor loops can be plotted. The
parameters of each loop can be obtained according to the experimental results which
will be presented in the following.
6.4 Results
The experimental setup of the proposed SMA actuator was the same as Fig. 3-5 in
chapter 3. The parameters of the experiments are listed in Table 6-1. To accurately
identify the parameters of the modified Liang and Rogers model formulated for
modeling the saturated hysteresis nonlinearity of an SMA actuator, the input voltage
applied to the SMA actuator in the training process is a slow decaying ramp signal,
making it possible to allow the temperature to stabilize, as in the steady state the
temperature will be decided by the applied voltage [62]. As shown in Fig. 6-7, the
slopes of the decay reversal curves are set to ±5.88×10-3
in the training process of the
modified Liang and Rogers model, including the maximum and minimum at 2.35V and
1.47V, respectively. Fig. 6-8 shows the results of the experiments with input
voltage-output displacement hysteresis loops for the SMA actuator using the input
voltage in Fig. 6-7. When the input voltage is below 1.47V or above 1.88V, the minor
hysteresis loops can be expressed by the modified Liang and Rogers model without the
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
119
increase stage. However, minor loops with input voltage inputV , 1.47V < inputV <1.88V,
where small changes in the input voltage lead to considerable displacement changes,
need to be expressed by the modified Liang and Rogers model with the increase stage.
Fig. 6-7 Schematic of the input voltage
Table 6-1 Parameters of the experiments
Ambient temperature 22°C SMA diameter 0.5mm
MOSFET K2232 SMA length 140mm
Power supply 5V Spring stiffness 653.3N/m
Microcontroller Arduino Pretension force 2.75N
0 200 400 600 800 10000
0.5
1
1.5
2
2.5
Volt
age
(V)
Time (s)
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
120
Then, parameter identification for the proposed model, simulation results by the
proposed model, and a comparison with the experimental data are presented. Estimates
for parameters used in Eq. (6-13) and Eq. (6-14) are obtained by curve fitting. The
remaining parameters are identified using the actual displacement measured by the laser
sensor, as actual temperature measurements are not available.
Based on Eq. (6-6) and Eq. (6-15), Fig. 6-9 shows the results of simulated austenite
fraction versus time, including the major and minor hysteresis loops of the SMA
actuator. The six modified Liang and Rogers model parameters are listed in Table 6-2.
minorsT , 2T , 3T , 4T , and minorfT are selected to identify the parameters in Eq. (6-10),
(6-12), (6-13), while minorsT , minorfT , and AFT are for the parameters in Eq. (6-14). The
corresponding major and minor loops of austenite fraction versus temperature are
shown in Fig. 6-10. It is evident that the hysteresis loops in Fig. 6-6 would qualitatively
match the hysteresis loops in Fig. 6-10. To obtain the martensite starting temperature
MSiT , Fig. 6-11 and Fig. 6-12 show the simulation results of MSiT based on the Eq.
(6-13) and Eq. (6-14), respectively. The results show that the martensite starting
temperature MSiT can be predicted.
With the relationship between input voltage and output temperature in Eq. (6-1) of
module 1, the input temperature and output austenite fraction in Eq. (6-6) and (6-15) of
module 2, the input austenite fraction and output displacement in Eq. (6-16) of module
3, as shown in Fig. 6-13, the input voltage VS output displacement with the modified
Liang and Rogers model is plotted, including the major loop plotted in a solid line and
the minor loops plotted in dashed lines.
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
121
Fig. 6-8 Schematic of the displacement VS input voltage
Fig. 6-9 Simulated austenite fraction versus time
0 0.5 1 1.5 2 2.5-0.2
0
0.2
0.4
0.6
0.8
1
0 200 400 600 8000
0.2
0.4
0.6
0.8
1Minor loopsMajor loop
Aust
enit
e fr
acti
on
Time (s)
Dis
pla
cem
ent
(mm
)
Voltage (V)
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
122
Fig. 6-10 Simulated austenite fraction versus temperature
Fig. 6-11 Curve for fitting the normalized martensite starting temperature
20 40 60 80 1000
0.2
0.4
0.6
0.8
1
Minor loops
Major loop
0 0.2 0.4 0.6 0.8 1-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Fitting curveNormalized temperature
Aust
enit
e fr
acti
on
Temperature (°C)
T1=Tminors
T2
T3
T4
T5=Tminorf
T6=TAF
No
rmal
ized
tem
per
ature
TM
Si (
)
Normalized temperature Ti (°C)
T1=Tminors
T2
T3
T4 T5=Tminorf
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
123
Fig. 6-12 Curve for fitting the martensite starting temperature
Fig. 6-13 Simulation results of the output displacement VS the input voltage
0.2 0.4 0.6 0.8 125
30
35
40
45
50
55
Fitting curveMartensite startTemperature
0 0.5 1 1.5 2-0.2
0
0.2
0.4
0.6
0.8
1
Major loop Minor loops
Tem
per
ature
TM
Si (°
C)
Austenite fraction )(i
Tahi
Tminors
Tminorf TAF
Dis
pla
cem
ent
(mm
)
Voltage (V)
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
124
Table 6-2 Simulation parameters
Parameter Value Parameter Value
6500kgm3 R 1Ω
h 165W/m2°C c 836.8 J/kg°C
q -32.69 e 60.47
AST 50.2°C AFT 80.1°C
MST 28°C MFT 22°C
MSfT 28.9°C MSsT 51.1°C
jf 0 js 0
ahf 0.96 ahs 0.29
Tf 75.9°C Ts 60.8°C
2MST 40.8°C 3MST 32.1°C
2j 0.16 3j 0.19
2ah 0.38 3ah 0.66
2T 63.1°C 3T 68.5°C
22T 61.2°C 33T 59.1
4MST 30.1°C 4T 71.5°C
4j 0.1 44T 65.2°C
4ah 0.82 g 0.96
4m 5.24 3m -12.80
2m 11.50 1m -4.94
0m 1.00
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
125
6.5 Consideration and Discussion
As shown in Fig. 6-14, the output with the modified Liang and Rogers model is
plotted together with the experimental data (colors other than red), with the major loop
plotted in a red solid line and the minor loops plotted in red dashed lines. This figure
clearly shows that the modified Liang and Rogers model can effectively characterize the
hysteresis behavior of the SMA actuator. However, there are some factors needed to be
discussed.
Fig. 6-14 Plot with experimental and simulated data
by the modified Liang and Rogers model
1. The Liang and Rogers model has found widespread use in SMA modeling and
position control, but the limitation has been mentioned above which the
temperature of SMA wire needs to be measured to decide the martensite and
0 0.5 1 1.5 2 2.5-0.2
0
0.2
0.4
0.6
0.8
1
MajorloopMinorloops
Others:Experimentaldata
Dis
pla
cem
ent
(mm
)
Voltage (V)
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
126
austenite phase. As shown in Fig. 6-13, the proposed model does not suffer from
this limitation since the simulation results show the relationship between input
voltage and output displacement. Concerning about the simulation and
experimental results, as shown in Fig. 6-14, six models are built to model the
basic shape of the hysteresis loops including major and minor loops. However,
the parameters (Table 6-2) of the model need to be changed once the ambient
conditions changed. For example, the variation of bias force leads to a shift of the
hysteresis. An increase in force causes a shift to the right, a decrease shifts the
hysteresis to the left [81], which leads to modify the parameters to fit the
experimental data. Therefore, it is difficult to use this model in unstable
environment.
2. As shown in Fig. 6-14, the discrepancy between the simulated and experimental
data can be explained as follows. The measured displacement shown in Fig. 6-8
corresponds to the voltage profile in Fig. 6-7. Thus, the output displacement
should be smooth. However, due to the ambient conditions of the experimental
setup are not constant over time and space, the error is unavoidable.
3. As shown in Fig. 6-8, the results of output displacement-voltage can repeat during
the heating process including the major and minor loops. However results during
the cooling process are different with since the martensite start temperature of
each loop is different. It is evident that the SMA actuator behavior is highly
nonlinear due to the complex physics.
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
127
6.6 Chapter Summary
In this chapter, a complete mathematical model for a spring-biased SMA actuator is
proposed, including both the major and minor hysteresis loops. With the difference from
reference [38] (Fig. 6-5), the proposed model is capable of simulating the temperature
(especially about the martensite start temperature of each loop), austenite fraction and
the output displacement of the actuator.
Based on the Liang and Rogers model, the modified Liang and Rogers model
including the major and minor hysteresis loops is developed and the model parameters
are adjusted by means of experimental identification procedure. The simulation results
show that good accuracy of the model can be assured by using the modified Liang and
Rogers model, including the increase and maintaining stages (expressed by Eq. (6-7)).
The proposed model can predict the martensite starting temperature of the minor
hysteresis loops, successfully.
Footnote: Junfeng Li, Hiroyuki Harada, Modeling of an SMA actuator based on the
Liang and Rogers model, International Journal of Applied Electromagnetics Mechanics,
Copyright (2013), accepted, with permission from IOS Press.
CHAPTER 6: MODELING OF AN SMA ACTUATOR BASED ON THE LIANG AND
ROGERS MODEL
128
CHAPTER 7: CONCLUSIONS AND FUTURE WORKS
129
CHAPTER 7:
CONCLUSIONS AND FUTURE WORKS
7.1 Conclusions
In many applications where space, weight and noise can be an issue, SMA actuators
present a potential solution with their high force-to-weight ratio, mechanical
compactness, ease of miniaturisation, as well as their clean and silent operation. With
improved speed, accuracy and controllability, the possibilities of using SMA actuators
in, to name a few, robotics, consumer appliances and bio-medical applications, are
greater now. In this thesis, the research focuses on achieving rapid control of SMA
actuators and the obtained results are documented and discussed. The main content of
this thesis can be summarized in the following.
In chapter 1, the objective, approach and the outline of thesis are introduced.
In chapter 2, literature overview which discusses the past research is presented in this
work, including modeling and control system design. Some research is verified by the
experimental results successfully. In addition, many methods developed using cooling
device, thin SMA wires and new actuator structures to achieve fast response speed are
introduced.
In chapter 3, the latency duration of the SMA wire is investigated. Experimental
results show that the latency duration affected by the ambient temperature, which is due
to the hysteresis effect can be observed.
In chapter 4, according to the definition of phase temperatures, the concept of phase
resistance is defined for the first time. Phase resistances used to divide the hysteresis
CHAPTER 7: CONCLUSIONS AND FUTURE WORKS
130
loop into four parts are applied to shorten the long latency duration of thick SMA wire.
The proposed method, phase resistance feedback control (PRFC), is presented here. The
critical aspect of the PRFC method is to shorten the total cooling time by controlling the
phase resistances. Two SMA wires connected by an insulation joint to prevent short
circuiting will cooperate with each other to verify the possibility of the proposed
methods. In order to show the advantages of the proposed method, the experimental
results are compared with the results of binary control. Since the latency duration of
heating process is shorter than cooling process, only the cooling process is considered in
the proposed method.
In chapter 5, phase resistance with displacement feedback control (PRDFC) is also
discussed, which shows another way to achieve rapid response speed and results are
compared with traditional displacement feedback control as well to demonstrate the
advantages of PRDFC. The central element of PRDFC is to divide the reference input
into two parts which are assigned to both SMA wires, separately. Since there is only one
displacement sensor to detect the position of SMA actuator, it is important to make sure
only one SMA wire shrinks or extends for each reference input part. In order to explain
the concept more clearly, a step signal is used as reference input. Combining the heating
and cooling processes of the two SMA wires, the latency durations of SMA wire is
shortened with the proposed method. And the results with PRDFC using ramp signal
also demonstrate that the proposed method achieves more rapid response speed than
traditional displacement feedback control.
In chapter 6, the complete mathematical model for a spring-biased SMA actuator is
proposed, including both the major and minor hysteresis loops. Based on the Liang and
Rogers model, the modified Liang and Rogers model including the major and minor
CHAPTER 7: CONCLUSIONS AND FUTURE WORKS
131
hysteresis loops is developed and the model parameters are adjusted by means of
experimental identification procedure.
To sum up, the research presented in this thesis represents a significant step forward
for practical SMA actuator applications in the future. Substantial improvements in terms
of speed of SMA actuator control have been made compared with the past work. All of
these have been accomplished with free convection cooling and not in a
temperature-controlled condition in speed and position control system.
7.2 Future Works
The present research is a step forward towards the aim of achieving rapid response
speed control of SMA actuator, and remains some limitations and insufficiency. In the
rapid response speed control system, the latency duration of SMA wire is short if the
diameter is small which makes it difficult to use phase resistance as feedback to obtain
rapid response speed by shortening the latency duration. For the concept of PRDFC
method, good results are demonstrated by the experimental tests. It is possible to test
this method using other different reference input, such as square wave.
Fig. 7-1 Block diagram for the compensation based on the inverse model
CHAPTER 7: CONCLUSIONS AND FUTURE WORKS
132
Due to the hysteresis effect of SMA, it is difficult to control the SMA actuator. To
completely compensate the hysteresis effect of an SMA system, it is important to
develop the exact inverse of the hysteresis model. Generally, as suggested in Fig. 7-1,
for the hysteresis model H and inverse model 1H , the following equation can be
developed if the inverse model exists [62].
uuHHvHy ))(()( 1 (7-1)
here y is the output displacement of modified Liang and Roger model; v is the output
voltage of the modified Liang and Roger inverse model which is used as the
feedforward in the proposed control system; u is the input displacement.
ACKNOWLEDGEMENTS
133
ACKNOWLEDGEMENTS
There are many people who had made contributions, both directly and indirectly,
towards the completion of this thesis. I would like to name a few. First and foremost, I
would like to express my sincerest gratitude to my supervisor, Dr. Hiroyuki Harada.
Harada sensei's help has been extremely enormous, and his enthusiasm has sparked my
own in this wonderful work. This work would never have been completed without his
advice and guidance.
I would also like to thank Prof. Kajiwara, Prof. Kobayashi and Prof. Nakamura who
give me many excellent advices for my research.
I would also like to express my gratitude to Dr. Werawan Manakul, the coordinator of
English Engineering Education Program, who supports my life and research in Sapporo.
I would also like to express my gratitude to my parents, who provided me with
unconditional advice, support and love. They have never been far from my heart.
134
APPENDIX A
135
APPENDIX A
a. Binary Control Code with Matlab
hold on
display('begin');
m=zeros(1,15);
n=0;
PP2=0;
PP3=0;
s = serial('COM5', 'BaudRate', 115200);
fopen(s);
pause(1);
% send data to arduino
fwrite(s,strcat(int2str(0),',',int2str(0),',',int2str(0)),'sync');
for i=1:2500
n=i
temp=fscanf(s); % get data from arduino
tempnum=str2num(temp);
m(i,1:length(tempnum))=tempnum;
% Displacement calculation
dis=5*m(:,5)/1023;
% Force calculation
f=5*m(:,15)/1023;
% Time calculation
APPENDIX A
136
t=m(i,4)/1000;
if t<5;
PP2(i)=0;
PP3(i)=0;
fwrite(s,strcat(int2str(0),',',int2str(PP2(i)),',',int2str(PP3(i))),'sync');
elseif t>5&t2<8
PP3(i)=0;
PP2(i)= 255;
fwrite(s,strcat(int2str(0),',',int2str(PP2(i)),',',int2str(PP3(i))),'sync');
else
PP2(i)=0;
PP3(i)=0;
fwrite(s,strcat(int2str(0),',',int2str(PP2(i)),',',int2str(PP3(i))),'sync');
end
end
display('done');
fclose(s);
delete(s);
APPENDIX A
137
b. PID Controller Code with Resistance as Feedback
kp=4000;
ki=0.2;
kd=0;
input2(i)= sF2(i);
setpoint2(i)=1.47; %resistance feedback
tlast1=0;
tchange=(now-tlast1);
error1(i)=setpoint1(i)-input1(i);
output1(i)=
output1+kp*(error1(i)-e1)+ki*error1(i)*tchange+kd*(error1(i)-2*e1+e2)/tchange;
if output2(i)>0
PP2(i)= 0;
elseif output2(i)<-255
PP2(i)=-255;
else
PP2(i)=-round(output2(i));
end
PP2(i)=- PP2(i); % PP2, 0-255, send to arduino to heat the SMA wire
tlast1=now;
e2=e1;
e1=error1(i);
output1= output1(i);
fwrite(s,strcat(int2str(0),',',int2str(PP2(i)),',',int2str(0)),'sync');
APPENDIX A
138
c. Microcontroller Code
Three PWM outputs and fifteen inputs are designed in this code which can be used to
test three segments of SMA wires. In the experiment, only two outputs are used.
String comdata = "";
int numdata[3] = 0, PWMPin[3] = 3, 4, 5, mark = 0;
int val;
int val1;
unsigned long time;
unsigned int x[3]= 0;
unsigned int y[3]= 0;
unsigned int z[3]= 0;
int analogin[9]= A0,A1,A2,A3,A4,A5,A6,A7,A8;
void setup()
Serial.begin(115200);
for(int i = 0; i < 3; i++) pinMode(PWMPin[i], OUTPUT);
void loop()
int j = 0;
while (Serial.available() > 0) // check to see if there is
data in the bus
APPENDIX A
139
comdata+=char(Serial.read());
delay(2);
val=analogRead(A11);
val1=analogRead(A15);
mark=1;
if(mark==1)
if(comdata.length() > 0)
for(int i = 0; i < comdata.length() ; i++)
if(comdata[i] == ',')
j++;
else
numdata[j] = numdata[j] * 10 + (comdata[i] - '0');
for(int i = 0; i < 3; i++)
digitalWrite(PWMPin[i],HIGH);
APPENDIX A
140
for(int j=0;j<3;j++)
x[i]+=analogRead(analogin[i]);
y[i]+=analogRead(analogin[i+3]);
z[i]+=analogRead(analogin[i+6]);
digitalWrite(PWMPin[i],LOW);
x[i]=(x[i]/3);
y[i]=(y[i]/3);
z[i]=(z[i]/3);
analogWrite(PWMPin[i],numdata[i]);
numdata[i] = 0;
Serial.print(x[0]);
Serial.print(",");
Serial.print(y[0]);
Serial.print(",");
Serial.print(z[0]);
Serial.print(",");
Serial.print(millis());
Serial.print(",");
Serial.print(val1);
Serial.print(",");
Serial.print(x[1]);
APPENDIX A
141
Serial.print(",");
Serial.print(y[1]);
Serial.print(",");
Serial.print(z[1]);
Serial.print(",");
Serial.print(x[2]);
Serial.print(",");
Serial.print(y[2]);
Serial.print(",");
Serial.print(z[2]);
Serial.print(",");
Serial.print(comdata);
Serial.print(",");
Serial.println(val);
mark = 0;
comdata = String("");
delay(10);
142
APPENDIX B
143
APPENDIX B
a. Microcontroller
In the control system, a microcontroller (Arduino Mega 2562, shown in Fig. B-1) is
used to collect the data from displacement sensor. The Arduino Mega 2560 is a
microcontroller board based on the ATmega2560 (datasheet). It has 54 digital
input/output pins (of which 15 can be used as PWM outputs), 16 analog inputs, 4
UARTs (hardware serial ports), a 16 MHz crystal oscillator, a USB connection, a power
jack, an ICSP header, and a reset button. It contains everything needed to support the
microcontroller; simply connect it to a computer with a USB cable or power it with a
AC-to-DC adapter or battery to get started. The Mega is compatible with most shields
designed for the Arduino Duemilanove or Diecimila. Detailed information is listed in
Table B-1.
Fig. B-1 Microcontroller used in experiment
APPENDIX B
144
Table B-1 Detailed information for microcontroller
Microcontroller ATmega2560
Operating Voltage 5V
Input Voltage 7-12V
Input Voltage (limits) 6-20V
Digital I/O Pins 54 (of which 15 provide PWM output)
Analog Input Pins 16
DC Current per I/O Pin 40 mA
DC Current for 3.3V Pin 50 mA
Flash Memory 256 KB of which 8 KB used by boot loader
SRAM 8 KB
EEPROM 4 KB
Clock Speed 16 MHz
b. Power Source, Displacement and Force Sensor
Since the SMA wire is heated by current, power source is supplied by TEXIO
Kenwood PA18-5B shown in Fig. B-2. The PA-B series is a high-performance DC
constant-voltage, constant-current power supply unit with 3.5 digit voltage indicator
LEDs and 3-digit current indicator LEDs. The series regulator control allows the user to
vary the output from 0 to the rated output. The output controller, a 10-turn winding type
variable resistor, offers fine control of output voltage and current. It is possible to set the
output voltage and current even the output is off. The output voltage and current may be
checked simultaneously. The PA-B series feature output On/Off control, output sensing,
APPENDIX B
145
and various remote control functions and fully meet various user needs. They have a
wide variety of applications including research and development, prototyping, test,
aging, and systems integration. The following is the features of TEXIO Kenwood
PA18-5B.
Fig. B-2 Power source used in experiment
Features for TEXIO Kenwood PA18-5B:
a. Low ripple, low noise
b. Digital display of voltage & current
c. Series/parallel operation
d. Floating output/voltage remote sensing terminal
e. External analog control for fine adjustments
f. EIA rack size
APPENDIX B
146
In order to detect the output displacement and force of SMA actuator, the
displacement and force are obtained by a KEYENCE LC-2000 laser displacement meter
(Fig. B-3)and a TEDEA-HUNTLEIGH load cell (Fig. 4), respectively. Since the output
displacement of SMA wire is less than 3 mm, the microcontroller can detect the output
voltage from displacement sensor directly when setting the displacement output voltage
from 0 to 3 V. However, the output voltage of load cell can be regulated shown in Fig.
B-4.
Fig. B-3 Displacement sensor used in experiment
Features for KEYENCE LC-2000:
a. Reference distance :40mm
b. Measurement range :±3mm
c. Analog output voltage :±3V
d. Resolution:0.5μm
e. Accuracy: ±10μm ±3%
APPENDIX B
147
Fig. B-4 Force sensor used in experiment
Features for TEDEA-HUNTLEIGH:
a. Capacity range: 5kg
b. Only 22mm high
c. Aluminum construction
d. Single point 350 x 350mm
e. IP66 protection
f. OIML R60 and NTEP approved
148
REFERENCES
149
REFERENCES
[1] Z. Wang, G. Hang, J. Li, Y. Wang and Kai Xiao, A micro-robot fish with embedded
SMA wire actuated flexible biomimetic fin, Sensors and Actuators A: Physical, Vol.
14, No. 2(2008), pp. 354–360.
[2] K. Ikuta, M. Tsukumoto and S. Hirose, Shape memory alloy servo actuator system
with electric resistance feedback and application for active endoscope, IEEE
International Conference on Robotics and Automation,Vol. 1, (1988), pp. 427-430.
[3] R. Featherstone and Y. H. Teh, Improving the speed of shape memory alloy
actuators by faster electrical heating, Experimental Robotics IX, Vol. 21, (2006), pp.
67-76.
[4] S. Vollach and D. Shilo, The mechanical response of shape memory alloys under a
rapid heating pulse, Experimental Mechanics, Vol. 50, No. 7(2010), pp. 803-811.
[5] B. Selden, K. J. Cho and H. H., Asada, Multi-segment state coordination for
reducing latency time of shape memory alloy actuator systems, Proceedings of the
2005 IEEE International Conference on Robotics and Automation, (2005), pp.
1350-1355.
[6] K. George and I. Mayo, "Memory Metal." Chem Matters, Vol. 4, No. 7(1993).
[7] B. Kiefer and D. C. Lagoudas, Magnetic field-induced martensitic variant
reorientation in magnetic shape memory alloys, Philosophical Magazine, Vol. 85,
No. 33-35(2005), pp. 4285-4329.
[8] I. Karaman, B. Basaran, H. E. Karaca, A. I. Karsilayan and Y. I. Chumlyakov,
Energy harvesting using martensite variant reorientation mechanism in NiMnGa
magnetic shape memory alloy, Applied physics letters, Vol. 90, No. 17(2007), pp.
172505-172505-3.
[9] H. Yu, Y. Kang, Z. Zhao, X. Wang and L. Chen, Microstructural characteristics and
texture of hot strip low carbon steel produced by flexible thin slab rolling with
warm rolling technology, Materials Characterization, Vol. 56, No. 2(2006), pp.
158-164.
[10] J. Yanagimoto and R. Izumi, Continuous electric resistance heating: hot forming
system for high-alloy metals with poor workability. Journal of Materials
REFERENCES
150
Processing Technology, Vol. 209, No. 6(2009), pp. 3060-3068.
[11] V. Brailovski, S. Prokoshkin, I. Khemelevskaya, K. Inaekyan, V. Demers, S.
Dobatkin and E. Tatyanin., Structure and properties of the Ti-50.0 % at Ni alloy
after strain hardening and nanocrystallizing thermomechanical processing,
Materials Transactions, Vol. 47, No. 3(2006), pp. 795-804.
[12] S. D. Prokoshkin, I. Yu. Khmelevskaya, S. V. Dobatkin, I. B. Trubitsyna, E. V.
Tatyanin, V. V. Stolyarov, E. A. Prokofiev, Alloy composition, deformation
temperature, pressure and post-deformation annealing effects in severely deformed
Ti–Ni based shape memory alloys. Acta Materialia, Vol. 53, No. 9(2005), pp.
2703-2714
[13] K. Ikuta, M. Tsukamoto, and S. Hirose, Mathematical model and experimental
verification of shape memory alloy for designing micro actuator, Proc. IEEE Micro
Electro Mechanical Systems (Piscataway, NJ: IEEE), (1991), pp. 103-108.
[14] J. D. Harrison, Measurable changes concomitant with the shape memory effect
transformation, Engineering Aspects of Shape Memory Alloys,
Butterworth-Heinemann, (1990), pp. 106-111.
[15] A. Nagasawa, K. Enami, Y. Ishino, Y. Abe, S. Nenno, Reversible shape memory
effect, Scripta Metallurgica, Vol. 8, No. 9(1974), pp. 1055-1060.
[16] T. Saburi and S. Nenno, Reversible shape memory in Cu-Zn-Ga, Scripta
Metallurgica, Vol. 8, No.12 (1974), pp. 1363-1367.
[17] V. Brailovski, S. Prokoshkin, P. Terriault, and F. Trochu, Shape memory Alloys:
Fundamentals, Modeling and Applications, University of Quebec higher institute of
technology, (2003).
[18] T. V. Duerig, K.N. Melton, D. Stockel and C. M. Wayman, Engineering aspects of
shape memory alloys, Chapter 1, Butterworth-Heinemann, (1990).
[19] K. Ikuta, Micro/miniature shape memory alloy actuator. In IEEE Robotics and
Automation Society, Vol. 3(1990), pp. 2156-2161.
[20] Toki Corporation, Tokyo- Japan, Biometal GlLidebook, (1987).
[21] K. Ikuta, M. Tsukamoto and S. Hirose, Proc. IEEE MEMS Workshop (1991) 108.
[22] O. K. Rediniotis, L. N. Wilson, D. C. Lagoudas and M. M. Khan, Development of a
shape-memory-alloy actuated biomimetic hydrofoil. Journal of Intelligent Material
Systems and Structures, Vol. 13, No. 1(2002), pp. 35-49.
REFERENCES
151
[23] A. David Johnson, V. Martynov, V. Gupta, Applications of shape memory alloys:
advantages, disadvantages, and limitations, Micromachining and Microfabrication,
International Society for Optics and Photonics, (2001), pp. 341-351.
[24] S. Suresh, Fatigue of Materials, Cambridge solid state science series, Cambridge
University Press, Cambridge, (1991).
[25] H. J. Christ, Wechselverformung von Metallen, Werkstoff-Forschung und-Technik
9, Springer, Berlin, (1991).
[26] H. O. Fuchs and R. I. Stephens, Metal Fatigue in Engineering, Wiley, New York,
(1980).
[27] J. J. Craig, Introduction to robotics: mechanics and control. Addison-Wesley
Publishing Group, 2nd edition, (1989).
[28] C. Liang and C.A. Rogers, Design of shape memory alloy actuators, J.intell.Mater.
Syst. &Struct, Vol. 6, No. 2(1995), pp. 220-338.
[29] K. Kuribayashi, A new actuator of a joint mechanism using Ti-Ni alloy wire. The
International Journal of Robotics Research, Vol. 4, No. 4(1986), pp. 47-58.
[30] K. Tanaka, A thermomechanical sketch of shape memory effect: One-dimensional
tensile behavior, Res Mechanica, International Journal of Structural Mechanics
and Materials Science, Vol. 18(1986), pp. 251-263.
[31] H. E. Mohammad and A. Hashem, Nonlinear control of a shape memory alloy
actuated manipulator, Journal of vibration and acoustics, Vol. 124, No. 4(2002), pp.
566-575.
[32] K. Tanaka and S. Nagaki, Thermomechanical description of materials with internal
variables in the process of phase transitions, Ingenieur-Archive, Vol. 51, No.
5(1982), pp. 287-299.
[33] J. G. Boyd and D. C. Lagoudas, A thermodynamic constitutive model for the shape
memory materials. Part I: the monolithic shape memory alloys, International
Journal of Plasticity, Vol. 12, No. 6(1996), pp. 805-842.
[34] C. Liang and C. A. Rogers, One-dimensional thermomechanical constitutive
relations for shape memory materials, Journal of Intelligent Material Systems and
Structures, Vol. 1, No. 2(1990), pp. 207-234.
[35] C. A. Rogers, C. Liang and C. R. Fuller, Modeling of shape memory alloy hybrid
composites for structural acoustic control, Journal of Acoustic Society of America,
REFERENCES
152
Vol. 89 No. 1(1991), pp. 210-220.
[36] W. S. Anders, C. A. Rogers, C. Liang and C. R. Fuller, Vibration and low-frequency
acoustic analysis of piecewise-activated adaptive composite panels, Journal of
Composite Materials, Vol. 26, No. 1(1992), pp. 103-120.
[37] D. Grant, Accurate and rapid control of shape memory alloy actuators. PhD thesis,
McGill University, (1999).
[38] S. M. Dutta and F. H. Ghorbel, Differential hysteresis modeling of a shape memory
alloy wire actuator, IEEE/ASME Transactions on Mechatronics Vol. 10, No.
2(2005), pp. 189-197.
[39] A. Visintin, Differential models of hysteresis. Berlin, Germany, Springer-Verlag,
(1994).
[40] J. Kopfová and P. Pavel Krejcí, A Preisach type model for temperature driven
hysteresis memory erasure in shape memory materials, Continuum Mechanics and
Thermodynamics, Vol. 23, No. 2(2011), pp. 125-137.
[41] R. Gorbet, D. Wang and K. Morris, Preisach model identification of a two-wire
SMA actuator, in Proceedings of IEEE International Conference on Robotics and
Automation, Vol. 3(1998), pp. 2161-2167.
[42] W. Galinaitis and R. Rogers, Compensation for hysteresis using bivariate Preisach
models. In SPIE Smart Structures and Materials, Vol. 3039(1997), pp. 538-547.
[43] O. Henze and W. M. Rucker, Identification procedures of Preisach model, IEEE
Trans. Magn., Vol. 38, No. 2(2002), pp. 833-836.
[44] E. Cardelli, E. D. Torre and G. Ban, Experimental determination of Preisach
distribution functions in magnetic cores, Phys. B: Cond. Matter, Vol. 275, No.
1(2000), pp. 262 -269.
[45] C. Natale, F. Velardi and C. Visone, Identification and compensation of Preisach
hysteresis models for magnetostrictive actuators, Phys.B:Cond.Matter, Vol. 306,
No. 1(2001), pp. 161-165.
[46] M. Ruderman and T. Bertram, Discrete dynamic Preisach model for robust inverse
control of hysteresis systems, Decision and Control (CDC), 2010 49th IEEE
Conference on. IEEE, (2010), pp. 3463-3468.
[47] R. Smith, Inverse compensation for hysteresis magnetostrictive transducers,
Mathematical and Computer Modelling, Vol. 33, No. 1(2001), pp. 285-298.
REFERENCES
153
[48] S. Bashash and N. Jalili, Robust multiple frequency trajectory tracking control of
piezoelectrically driven micro/nanopositioning systems, IEEE Trans Control Syst
Technol, Vol. 15, No. 5(2007), pp. 867-878.
[49] P. Krejci, Hysteresis, convexity and dissipation in hyperbolic equations,
Gakkotosho, Tokyo, (1996).
[50] P. Krejci and K. Kuhnen, Inverse control of systems with hysteresis and creep,
Proc. Inst. Elect. Eng.-Control Theory Appl., Vol. 148, No. 3(2001), pp. 185-192.
[51] X. Chen and T. Hisayama, Adaptive sliding-mode position control for
piezo-actuated stage, IEEE Trans. Ind. Electron., Vol. 55, No. 11(2008), pp.
3927-3934.
[52] K. Kuhnen, Modeling, Identification and compensation of complex hysteretic
nonlinearities, Eur. J. Control, Vol. 9(2003), pp. 407-418.
[53] M. Brokate and J. Sprekels, Hysteresis and phase transitions, Springer, New York,
(1996).
[54] A. Visintin, Differential models of hysteresis, Springer, Berlin, (1994).
[55] P. Krejci and K. Kuhnen, Inverse control of systems with hysteresis and creep,
IEEE Proc Control Theory Appl, Vol. 148, No. 3(2001), pp. 185-192.
[56] P. Ge and M. Jouaneh, Tracking control of a piezoceramic actuator, IEEE Trans.
Control Syst. Technol., Vol. 4, No. 3(1996), pp. 209-216.
[57] J. Nealis and R. C. Smith, Model-based robust control design for magnetostrictive
transducers operating in hysteretic and nonlinear regimes, IEEE Trans. Control
Syst. Technol., Vol. 15, No. 1(2007), pp. 22-39.
[58] C. Ru, L. Chen, B. Shao, W. Rong and L. Sun, A hysteresis compensation method
of piezoelectric actuator: model, identification and control, Control Eng Pract, Vol.
17, (2009), pp. 1107-1114.
[59] X. Tan and J. S. Baras, Adaptive Identification and Control of Hysteresis in Smart
Materials, IEEE Transactions on Automatic Control, Vol. 50, No. 6(2005), pp.
827-839.
[60] G. Song, J. Zhao, X. Zhou and J. Abreu-Garcia, Tracking control of a piezoceramic
actuator with hysteresis compensation using inverse Preisach model, IEEE/ASME
Trans. Mechatronics, Vol. 10, No. 2(2005), pp. 198-209.
[61] M. A. Janaideh, S. Rakheja and C. Y. Su, An analytical generalized
REFERENCES
154
Prandtl-Ishlinskii model inversion for hysteresis compensation in micropositioning
control, IEEE/ASME Trans Mech, Vol. 16, No. 4(2011), pp. 734-744.
[62] S. Hassan and R. Z. Mohammad, Position control of shape memory alloy
actuator based on the generalized Prandtl-Ishlinskii inverse model, Mechatronics,
Vol. 22, No. 7(2012), pp. 945-957.
[63] E. Asua, V. Etxebarria and A. Garcia-Arribas, Micropositioning control using shape
memory alloys, In: Proceedings of IEEE conference on control applications
CCA2006, Munich, Germany, (2006), pp. 3229-3234.
[64] E. P. Da Silva, Beam shape feedback control by means of a shape memory actuator,
Materials & design, Vol. 28, No. 5(2007), pp. 1592-1596.
[65] A.V. Popov, M. Labib, J. Fays and R. M. Botez, Closed-loop control simulations on
a morphing wing, Journal of Aircraft, Vol. 45, No. 5(2008), pp. 1794-1803.
[66] E. Shameli, A. Alasty and H. Salaarieh, Stability analysis and nonlinearity control
of a miniature shape memory alloy actuator for precise applications, Mechatronics,
Vol. 15, No. 4(2005), pp. 471-486.
[67] K. Ahn and B. Nguyen, Position control of shape memory alloy actuators using
self-tuning fuzzy PID controller, Int J Control, Automat, Syst, Vol. 4, No. 6(2006),
pp. 756-762.
[68] J. Carvajal, G. Chen and H. Ogmen, Fuzzy PID controller: design, performance
evaluation, and stability analysis, Inform Sci, Vol. 123, No. 3(2000), pp. 249-270.
[69] H. Li and H. Gatland, A new methodology for designing a fuzzy logic controller,
IEEE Transactions on, Vol. 25, No. 3(1995), pp.505-512.
[70] C. Cocaud, A. Price, A. Jnifene and H. Naguib, Position control of an experimental
robotic arm driven by artificial muscles based on shape memory alloys,
International Journal of Mechanics and Materials in Design, Vol. 3, No. 3(2006),
pp. 223-236.
[71] N. Ma and G. Song, Control of shape memory alloy actuator using pulse width
modulation, Smart Mater. Struct., Vol. 12, No. 5(2003), pp. 712-719.
[72] N. Ma and G. Song, Control of shape memory alloy actuators using pulse-width
pulse-frequency (PWPF), J Intell Mater Syst Struct, Vol. 14, No. 1(2003), pp.
15-22.
[73] J. Ko, M. B. Jun, G. Gilardi, E. Haslam and E. J. Park, Fuzzy PWM-PID control of
REFERENCES
155
cocontracting antagonistic shape memory alloy muscle pairs in an artificial finger,
Mechatronics, Vol. 21, No. 7(2011), pp. 1190-1202.
[74] D. Grant and V. Hayward, Variable structure control of shape memory alloy
actuators, IEEE Control Syst Mag, Vol. 17, No. 3(1997), pp. 80-88.
[75] J. Jayender, R.V. Patel, S. Nikumb, M. Ostojic, Modelling and gain scheduled
control of shape memory alloy actuators, Control Applications, 2005. CCA 2005.
Proceedings of 2005 IEEE Conference on. IEEE, (2005), pp. 767-772.
[76] A. Kumagai, P. Hozian and M. Kirkland, Neuro-fuzzy model based feedback
controller for shape memory alloy actuators, Proc. SPIE, (2000), pp. 291-299.
[77] G. Song, V. Chaudhry and C. Batur, Precision tracking control of shape memory
alloy actuators using neural networks and a sliding-mode based robust controller,
Smart Materials and Structures, Vol. 12, No. 2(2003), pp. 223-231.
[78] G. V. Webb and D. C. Lagoudas, Hysteresis modeling of SMA actuators for control
application, Journal of Intelligent Material Systems and Structures, Vol. 9, No.
6(1998), pp. 432-448.
[79] G. Webb, L. Wilson, D. Lagoudas, and O. Rediniotis, Adaptive control of shape
memory alloy actuators for underwater biomimetic applications, AIAA journal, Vol.
38, No. 2(2000), pp. 325-334.
[80] Y. H. Teh and R. Featherstone, An architecture for fast and accurate control of
shape memory alloy actuators, Int. J. Robotics Research, Vol. 27, No.5(2008),
pp.595-611.
[81] B. Selden, K. J. Cho and H. H. Asada, Segmented shape memory alloy actuators
using hysteresis loop control, Smart Mater Struct, Vol. 15, No. 2(2006), pp.
624-625..
[82] T. Hasegawa and S. Majima, A control system to compensate the hysteresis by
Preisach model on SMA actuator, Proceedings of the 1998 International
Symposium on Micromechatronics and Human Science, (1998), pp. 171-176.
[83] C. C. Lan, Investigation on pre-tensioned shape memory alloy actuators for force
and displacement self-sensing, in Intelligent Robots and Systems (IROS), IEEE/RSJ
International Conference, Taiwan, (2010), pp. 3043-3048.
[84] E. Asua, J. Feutchwanger, A. García-Arribas, V. Etxebarria, Sensorless control of
SMA-based actuators using neural networks, Journal of Intelligent Material
REFERENCES
156
Systems and Structures [J. Intellig. Mater. Syst. Struct.], Vol. 21, No. 18(2010), pp.
1809-1818.
[85] N. Ma, G. Song and H. Lee, Position control of shape memory alloy actuators with
internal electrical resistance feedback using neural networks, Smart Mater. Struct.,
Vol. 13,No. 4(2004), pp. 777-783.
[86] S. Hirose, K. Ikuta and Y. Umetani, A new design method of servo-actuators based
on the shape memory effect, Proc. of 5th RO.MAN.SY.symp. (1984); pp. 339-349.
[87] E. Williams and H. E. Mohammad, An Automotive SMA mirror actuator: modeling,
design, and experimental evaluation, Journal of Intelligent Material Systems and
Structures, Vol. 19, No. 12(2008), pp.1425-1434.